Download as pdf or txt
Download as pdf or txt
You are on page 1of 18

Additive Manufacturing 35 (2020) 101291

Contents lists available at ScienceDirect

Additive Manufacturing
journal homepage: www.elsevier.com/locate/addma

Comparison of ranking models to evaluate desktop 3D printers in a growing T


market
Amit J. Lopes*, Mireya A. Perez, David Espalin, Ryan B. Wicker
W.M. Keck Center for 3D Innovation, The University of Texas at El Paso, United States

ARTICLE INFO ABSTRACT

Keywords: While additive manufacturing (AM), commonly known as 3D printing, has been in existence commercially for
Desktop 3D printers ∼30 years, desktop 3D printers are a relatively new and rapidly growing market segment. Both well-established
Additive manufacturing AM companies and an increasing number of new enterprises are designing and retailing desktop systems of
Ranking models various sizes, capabilities, and prices. With the abundance of desktop systems now on the market, a consumer
may benefit from determining which system best serves their needs. This research highlights differences amongst
45 desktop 3D printers and suggests a method by which to evaluate such differences. For this, a standard part
consisting of various geometric features was designed and printed using each system. An updated version of a
previously developed quantitative ranking model was utilized to rate the build precision of each system as well
as other features, including build volume, size, cost, weight, and layer resolution. In addition, the research team
observed part aesthetics and quantified mechanical properties. The criteria evaluated in this ranking model may
be modified by each user, to extend this methodology to other desktop AM systems, including professional-grade
machines. The results from the model presented in this research were compared with other commonly used
ranking methods to help evaluate each technique. These included a simple ascending order rank based model
adjusted for ties (1-best and 45-worst), a percentile value based model evaluating the factor contribution values
presented in this paper; and the Technique for Order Preference by Similarity to an Ideal Solution (TOPSIS)
which evaluated separation measures from ideal and worst solutions. As expected, the comparisons demon-
strated that each model had slightly different rankings as compared to the model presented in this paper, with
some outliers. Consequently, it was observed that percentile value based models (such as one presented in this
paper) provide rankings different from separation value based models (from ideal and worst solutions) such as
TOPSIS.

1. Introduction upsurge in desktop system market availability. As per the Wohlers 2018
Report, the AM industry, consisting of all AM products and services
Additive manufacturing (AM) is the process of producing a com- worldwide, exceeded $7.3 billion in the year 2017 [2]. A recent (April
puter-designed three-dimensional (3D) object through addition of ma- 2020) search of ‘3D printers’ on crowdfunding platforms, like Kick-
terial in a layerwise method. Since Chuck Hull invented stereo- starter and Indiegogo lists over 200 companies promoting 3D printing
lithography, the first commercialized 3D printing process, in the 1980s, capabilities across the two platforms. In 2014, 278,000 desktop systems
the technology has experienced tremendous advances [1]. Although were purchased, while in 2017, an estimated 528,952 were purchased,
AM was used mostly in academic research or large-scale prototyping nearly doubling the number in just three years. While in 2014, desktop
commercial settings, it has since expanded to full scale production parts printer sale profits increased to $173.5 million, they were $500 million
worldwide (see as examples, GE Aviation, General Motors, 3DEO, Re- by 2017, a nearly threefold increase [2,3].
solution Medical). Today, the demand for smaller, more affordable 3D The average selling price of a professional-grade AM system in 2014
printers has grown amongst hobbyists, do-it-yourselfers, design en- was $87,140. However, this number was the result of a market that was
gineers, and small business owners. 3D printing has also increased in dominated by polymer-based systems. Metal AM systems, which are
popularity amongst teachers and professors using the technology for typically more expensive than polymer AM systems are increasing in
educational purposes. As a result, there is a seemingly continuous sales on an annual basis. For example, Arcam (Mölndal, Sweden,


Corresponding author.
E-mail address: ajlopes@utep.edu (A.J. Lopes).

https://doi.org/10.1016/j.addma.2020.101291
Received 22 January 2020; Received in revised form 23 April 2020; Accepted 26 April 2020
Available online 11 May 2020
2214-8604/ © 2020 Elsevier B.V. All rights reserved.
A.J. Lopes, et al. Additive Manufacturing 35 (2020) 101291

acquired by GE beginning in 2016), sold 24 systems in 2012, but saw an dimensional accuracy, overhangs, tolerances, bridging, and other cri-
increase to 42 systems in less than two years. Their Q20 powder bed teria [14]. In 2013, Roberson et al. used a benchmark part to rank
fusion metal machine retailed at a base price of €800,000 (∼ $896,000) various desktop systems according to cost, build time, material usage,
in 2015. On the other hand, the Fortus 900mc (Stratasys, Eden Prairie, and geometric accuracy [15]. They developed a ranking model that
MN, USA), Stratasys’ most expensive material extrusion system, had a used these factors as inputs to provide each system with a score. One of
base price of $400,000 in 2015 [3]. One-hundred and thirty-five com- the benefits of the model is that it was structured in a way that allowed
panies around the world produced and sold industrial AM systems in modifications, such as changing the factors being evaluated or the
2017 compared to 97 companies in 2016. The increase in popularity of number of systems being considered. However, their work only tested
metal AM is evident, as 1,768 metal AM systems were sold in 2017, five systems.
compared to 983 systems in 2016, an increase of nearly 80 percent [2]. The objective of the research presented here was to build upon
One of the first 3D printers advertised as “low-cost” was the previous work by Roberson et al. by using a new test part; one that
Dimension 3D machine, released by Stratasys in 2002. Priced at ap- included a wider array of features such as ramps for measuring surface
proximately $30,000, it was hardly affordable for the average user [4]. roughness, a tensile bar for conducting mechanical tests, and others to
In 2007, Adrian Bowyer and his team created the RepRap project and observe part aesthetics [16]. This new design was derived from parts
built the RepRap Version I Darwin. The idea behind this project was the previously designed in [9–15], especially that of Moylan et al. as their
creation of a self-replicating open-source system, one that can print out design seemed to capture features that tested system performance well
a significant portion of its own parts, while the remaining parts can be [13]. Using this updated test artifact, 45 systems were evaluated and
purchased off-the-shelf [5]. In fact, three of the systems tested in this ranked according to various factors. It should be noted the materials
paper, the Mendel, the Prusa Mendel, and the Mini Kossel, are succes- and machines used in the ranking model were circa 2015, corre-
sors of the Darwin system. Open source printing has increased in po- sponding to when the test artifacts from the 45 systems were produced.
pularity [6,7] and many of the top desktop printer manufacturers today Further, the methods described here can be continuously updated,
are still using this concept in the construction of their printers. In 2009, improved and used for evaluating contemporary systems as new ma-
MakerBot released the Cupcake, one of the first commercially available chines, features, capabilities, and materials enter the market and im-
desktop 3D printers. In 2013, Stratasys acquired MakerBot elevating the provements to the ranking model become available. In the ranking
importance of the desktop 3D printer market. Each year, the number of model, the following factors were taken into consideration: system
desktop 3D printer manufacturers and their market continuously ex- build volume, system size, system weight, system cost, layer resolution,
pands as many existing and new companies release printers with dimensional deviations, surface roughness, and positional error along
greater capabilities and better features. the x and y-axes. Finally, the results from the ranking model were
A significant majority of desktop 3D printers use material extrusion compared with three commonly used ranking strategies: a simple as-
technology, as a result of the expiration of early AM patents, specifically cending order rank based model adjusted for ties (1-best and 45-worst),
fused deposition modeling (FDM) patents owned by Stratasys. Other a percentile value based model evaluating the factor contribution va-
desktop systems employ the vat photopolymerization technology, a lues presented in this paper; and the Technique for Order Preference by
result of the expiration of 3D Systems’ patents, the original stereo- Similarity to an Ideal Solution (TOPSIS) which evaluated separation
lithography machine manufacturers. One of the systems evaluated in measures from the ideal and worst solutions.
this work uses the sheet lamination process, where sheets of material
are bonded to form an object. The recent expiration of selective laser 2. Systems evaluated
sintering (SLS) patents is also leading to increased market availability of
desktop printers that use powder bed fusion technology. These and The 45 desktop systems evaluated in this paper are listed in Table 1
other AM terms are defined as per ISO/ASTM 52900 [8]. and were chosen based on their accessibility at the time the research
As with any other purchase, when a consumer acquires a desktop 3D study was conducted. Of the systems tested, some are no longer in
printer, they expect to receive a high quality system that prints an production by the manufacturers as they have been discontinued or
identical replica of the CAD design that has been created. Researchers have been replaced by newer generations (indicated by a superscript 1).
have previously used benchmark parts, to rank AM systems and tech- The authors decided to keep these systems in the ranking as some of
nologies. In 1994, Childs et al. developed a test part that included free- them are still sold via other retailers, such as Amazon, eBay, and others.
form features, such as fish-shaped and handle-like structures, to eval- It is also important to note that Stratasys and 3D Systems have since
uate six AM systems within four different technologies [9]. In 2000, acquired MakerBot and Bits from Bytes, respectively. Also, competition
Zhou et al. tested various vat photopolymerization systems with a test within the desktop 3D printer industry is increasing at a very rapid rate
part that included round, square, and triangular features used to test and this study has the potential to be repeated on an annual basis for
horizontal and vertical dimensions, roundness, sphericity, angularity, evaluating the most current systems.
flatness, and surface roughness [10]. In 2004, Mahesh et al. used a
benchmark part to evaluate four AM polymer technologies for geo- 2.1. System specifications
metric accuracy, warpage, and surface roughness [11]. In 2009, Espalin
et al. designed a part to test how vapor smoothing improves parts built System specifications for all 45 systems are provided in Table 2. It is
by material extrusion [12]. In 2012, Moylan et al. proposed a part important to note that these are manufacturer advertised specifications,
design which included orifices, ramps, staircases, flat surfaces, and not values measured as part of these tests. Lightweight systems, like the
lateral features to evaluate both, polymer and metal processes [13]. 3D Systems - Cube, Afinia - H480, RepRap - Mini Kossel, and Formlabs -
While this is not an all-inclusive list of previous work in this field, it Form 1+, which all weigh under 10 kg, are well-suited for educators
provides a good understanding of how authors design parts to test and who may benefit from their portability. These systems are also good for
compare AM systems and technologies. at-home users and those working in remote-locations. Typically,
It is important to note that much of the work described above was smaller systems like these have very small build volumes (< 3,000
conducted using professional-grade AM systems which use alkali-so- cm3). However, the Mini Kossel from RepRap is an exception, having a
luble thermoplastic support materials and have temperature controlled 4,767 cm3 build volume while weighing only 5.5 kg. This is a delta type
heated chambers. Some research has taken place to test the efficiency printer in which the extrusion head is moved by three arms while the
and accuracy of desktop printers. Make Magazine, in the past, has build platform is stationary. Heavier systems like the 3D Touch and BFB
conducted a very thorough annual evaluation of some of the desktop 3000 from Bits from Bytes, or the MakerBot - Replicator Z18, Solido -
printers on the market, using a variety of prints that test for SD300 Pro, and Stratasys - uPrint Plus, which all weigh above 35 kg,

2
A.J. Lopes, et al. Additive Manufacturing 35 (2020) 101291

Table 1
Evaluated Systems.
Manufacturer System AM Technology Headquarters

1 st
3D Systems Cube (1 Generation) Material Extrusion Rock Hill, SC, USA
Afinia H4801 Material Extrusion Chanhassen, MN, USA
Airwolf 3D HD2X Material Extrusion Costa Mesa, CA, USA
BeeVeryCreative BeeTheFirst1 Material Extrusion Ílhavo, Portugal
Bits from Bytes2 3D Touch1 Material Extrusion Bristol, UK
Bits from Bytes2 BFB 30001 Material Extrusion Bristol, UK
Bits from Bytes2 RapMan 3.11 Material Extrusion Bristol, UK
Builder 3D Printers Builder Dual Feed1 Material Extrusion Noordwijkerhout, The Netherlands
Builder 3D Printers Big Builder Dual Feed1 Material Extrusion Noordwijkerhout, The Netherlands
CEL Robox1 Material Extrusion Bristol, UK
Craft Unique CraftBot1 Material Extrusion Budapest, Hungary
Dremel 3D Idea Builder1 Material Extrusion Mount Prospect, IL, USA
Felix Felix 3.01 Material Extrusion De Meern, The Netherlands
Flashforge Creator1 Material Extrusion Jinhua, China
Flashforge Dreamer Material Extrusion Jinhua, China
Formlabs Form1+1 Vat Photopolymerization Somerville, MA, USA
FSL3D Pegasus Touch Vat Photopolymerization Las Vegas, NV, USA
Leapfrog Creatr HS1 Material Extrusion Alphen aan den Rijn, The Netherlands
LulzBot Mini Material Extrusion Loveland, CO, USA
LulzBot TAZ 41 Material Extrusion Loveland, CO, USA
MakerBot3 Replicator1 (1st Generation) Material Extrusion Brooklyn, NY, USA
MakerBot3 Replicator1 (5th Generation) Material Extrusion Brooklyn, NY, USA
MakerBot3 Replicator 21 Material Extrusion Brooklyn, NY, USA
MakerBot3 Replicator 2X1 Material Extrusion Brooklyn, NY, USA
MakerBot3 Replicator Mini1 Material Extrusion Brooklyn, NY, USA
MakerBot3 Replicator Z18 Material Extrusion Brooklyn, NY, USA
MakerBot3 Thing-O-Matic1 Material Extrusion Brooklyn, NY, USA
Markforged Mark One1 Material Extrusion Somerville, MA, USA
MBot Cube1 Material Extrusion Hangzhou, China
PP3DP Up! Plus1 (1st Generation) Material Extrusion Beijing, China
Printrbot Metal Plus1 Material Extrusion Lincoln, CA, USA
Printrbot Simple Metal1 Material Extrusion Lincoln, CA, USA
Rapide 3D Rapide Lite 2001 Material Extrusion Shenzhen, China
RepRap Mendel1 Material Extrusion Bath, UK
RepRap Mini Kossel1 Material Extrusion Bath, UK
RepRap Prusa Mendel1 (1st Iteration) Material Extrusion Bath, UK
SeeMeCNC Rostock Max1 Material Extrusion Goshen, IN, USA
Solido SD300 Pro1 Sheet Lamination Manchester, NH, USA
Solidoodle Solidoodle 41 Material Extrusion Brooklyn, NY, USA
Stratasys Mojo1 Material Extrusion Eden Prairie, MN, USA
Stratasys uPrint Plus1 Material Extrusion Eden Prairie, MN, USA
Type A Machines Series 11 Material Extrusion San Francisco, CA, USA
Ultimaker Ultimaker 21 Material Extrusion Geldermalsen, The Netherlands
XYZ Printing Da Vinci 2.0 Duo Material Extrusion New Taipei City, Taiwan
Zortrax M2001 Material Extrusion Olsztyn, Poland

1 2 3
System Discontinued, Acquired by 3D Systems, Acquired by Stratasys.

work best in stationary locations and may be more useful in small- However, the above limitations will be addressed to include com-
business settings. While, the build volume in the 3D Touch and BFB mercial AM systems in future research. Some additional parameters that
3000 benefits from their large size and weight, the uPrint Plus has a will need to be considered to enable their inclusion are maintenance
smaller build volume than some of the other systems which weigh much costs, facilities upgrades, space constraints, raw materials costs, pre-
less. This increase in weight can be largely attributed to the additional and post-processing equipment costs, software costs, energy costs, and
equipment included in the uPrint Plus (heated build envelope, material more.
bays, and convection fans). Some of the systems tested include an enclosed build envelope (i.e.,
Some may argue the uPrint Plus should not be considered in these chamber surrounding the build platform). This is a safety feature to
tests because of its high price point of $19,900 (circa 2015). However, it protect the user from moving parts during a build. A superscript “C” in
was included because it is much more compact than other commercial the Envelope/Build Volume column designates systems with an en-
AM systems and serves to provide a good comparison to more sophis- closed envelope in Table 2. The build envelopes of the Stratasys - Mojo,
ticated systems. It is important to keep in mind that the system prices uPrint Plus, and MakerBot - Replicator Z18 are also temperature-con-
provided here have changed and are changing as new products become trolled, which helps prevent the part from shrinking or warping [17],
available. The researchers acknowledge the need to include commercial and improves interlayer bonding [18]. This feature is not common in
printers in such studies, but focused on making advancements toward other desktop printers because the patent is still currently owned by
understanding the ranking models for this specific effort. Some of the Stratasys [17]. The Solido - SD300 Pro’s build envelope builds at a
reasons to not include commercial printers in this research include: temperature of 35 °C, slightly higher than room temperature. Other
systems, come standard with a heated build platform, designated by an
• Increased cost to acquire parts “H,” that serves a similar purpose as a heated envelope. Heated build
• Limited and difficult access to many commercial AM systems platforms for some of the other systems are available for purchase as
• Lack of standardized benchmark part for long-term validation of add-on accessories. The best layer resolution advertised (in 2015) was
system capability for the Formlabs - Form1+ and FSL3D - Pegasus Touch at 25 μm, as

3
A.J. Lopes, et al. Additive Manufacturing 35 (2020) 101291

Table 2
System Specifications (circa, 2015).
System System Dimensions System Weight Envelope/ System Cost Highest Layer Resolution Source
Build Volume
mm kg cm3 USD μm

H
3D Systems - Cube 260 × 260 × 340 4.3 2744 $699 250 [20]
Afinia - H480 245 × 260 × 350 5 2646H $1299 200 [21]
Airwolf 3D - HD2X 400 × 140 × 400 18 18,878H $3995 60 [22]
BeeVeryCreative - BeeTheFirst 515 × 515 × 598 9.5 3206 $1699 50 [23]
Bits from Bytes - 3D Touch 515 × 515 × 598 36 15,881 $3930 125 [24]
Bits from Bytes - BFB 3000 515 × 515 × 590 36 15,881 $3300 125 [25]
Bits from Bytes - RapMan 3.1 570 × 650 × 820 17 11,623 $1930 125 [26]
Builder 3D Printers - Builder Dual Feed 385 × 370 × 400 15 7577 $1878 50 [27]
Builder 3D Printers - Big Builder Dual Feed 385 × 370 × 900 25 30,677C $2798 50 [28]
CEL - Robox 370 × 340 × 240 8 3150CH $1799 20 [29]
Craft Unique - CraftBot 400 × 460 × 450 14 10,000H $870 100 [30]
Dremel 3D - Idea Builder 400 × 485 × 335 8.8 4830 $999 100 [31]
Felix - Felix 3.0 450 × 500 × 350 8.7 11,762H $1499 50 [32]
Flashforge - Creator 320 × 467 × 381 12 4916CH $977 100 [33]
Flashforge - Dreamer 320 × 467 × 381 12 4851CH $1299 100 [34]
Formlabs - Form1+ 300 × 280 × 450 8 2578C $3299 25 [35]
FSL3D - Pegasus Touch 279 × 356 × 572 13 7143C $2999 25 [36]
Leapfrog - Creatr HS 600 × 500 × 500 32 12,528H $2241 20 [37]
LulzBot - Mini 435 × 340 × 385 8.6 3650H $1350 50 [38]
LulzBot - TAZ 4 680 × 520 × 515 11 20,488H $2195 75 [39]
MakerBot - Replicator (1st Generation) 320 × 467 × 381 14.5 4894H $1999 200 [40]
MakerBot Replicator (5th Gen.) 528 × 441 × 410 16 7522 $2899 100 [40]
MakerBot - Replicator 2 490 × 420 × 380 11.5 6759 $2199 100 [41]
MakerBot - Replicator 2X 490 × 320 × 531 12.6 5796CH $2499 100 [42]
MakerBot - Replicator Mini 295 × 310 × 381 8 1250C $1375 200 [43]
MakerBot - Replicator Z18 493 × 565 × 861 41 41,816C $6499 100 [44]
MakerBot - Thing-O-Matic 300 × 300 × 400 10.4 1452H $1225 300 [45]
Markforged - Mark One 575 × 322 × 360 22.7 6758C $5499 100 [46]
MBot - Cube 405 × 405 × 410 15 9350 $799 100 [47]
PP3DP - Up! Plus 245 × 260 × 350 5 2646 $1499H 150 [48]
Printrbot - Metal Plus 610 × 610 × 610 9.1 16,387H $1199 100 [49]
Printrbot - Simple Metal 457 × 432 × 330 7.3 3375 $599 100 [50]
Rapide 3D - Rapide Lite 200 360 × 350 × 380 10 8000H $999 50 [51]
RepRap - Mendel 400 × 500 × 360 7 5600 $520 300 [52]
RepRap - Mini Kossel 300 × 600 5.5 4767 $600 30 [53]
RepRap - Prusa Mendel 400 × 500 × 360 7 4000 $599 300 [54]
SeeMeCNC - Rostock Max 457 × 1219 11 19,581H $999 100 [55]
Solido - SD300 Pro 770 × 460 × 420 36 4536C $4375 168 [56]
Solidoodle - Solidoodle 4 343 × 356 × 381 12.7 8365CH $699 100 [57]
Stratasys - Mojo 630 × 450 × 530 27 2048C $5900 170 [58]
Stratasys - uPrint Plus 660 × 635 × 787 76 6264C $19,900 250 [59]
Type A Machines - Series 1 766 × 572 × 459 13.6 28,373 $2749 50 [60]
Ultimaker - Ultimaker 2 357 × 342 × 388 11.2 10,194H $2124 20 [61]
XYZ Printing - Da Vinci 2.0 Duo 468 × 510 × 558 27.5 6000CH $649 100 [62]
Zortrax - M200 345 × 430 × 430 13 7400H $1990 25 [63]

C H
System has enclosed build envelope, System has heated build platform.

well as the CEL - Robox, Leapfrog - Creatr HS, and the Ultimaker 2 at 20 of each protruding ridge while the negative ridges (feature C) were used
μm. Again, these are manufacturer provided specifications and the ac- to measure the open spaces in-between. The positive and negative
curacy at various print resolutions was not compared . descending cylinders (features D and E) were used to measure varying
diameters. Each step of the staircases (features F and G) was used to
obtain various values of height. The cylinders, both positive and ne-
3. Materials and methods gative (features H and I), were used to obtain the diameter of a re-
peating identical feature. Features J-L were used to measure other
3.1. Test part design factors as noted in the table. The four ramps (feature J) were used to
quantify surface roughness along planes inclined at 10°, 15°, 30°, and
After a literature review on test parts for AM systems was con- 45° as has been done in other work [13,15,64]. Positional error was
ducted, a final test part design was agreed upon for this research. This determined by using the rectangular prisms (feature K). The tensile bar
design, seen in Fig. 1, included shapes and features that provided im- (feature L) was used to measure ultimate tensile strength of the model
portant information regarding the capabilities and limitations of each material. Certain features were only used to inspect the visual aesthetics
desktop system. of the part, including the thin angled wall (feature M), positive and
The list of features is provided in Table 3. Positive features are those negative domes (features N and O), thin cylindrical wall (feature P),
protruding above the surface of the rectangular base while negative and positive and negative descending rectangular prisms (features Q
features are those occurring below the surface. Dimensional deviations and R).
were tested with features A-J. Although ultimate tensile strength and visual aesthetics were not
The square base (feature A) was used to measure the thickness of the included in the ranking model, the authors felt it was important to
part, which was dependent upon the accuracy of each layer thickness. include the results in this paper. Other authors have suggested
The positive lateral ridges (feature B) were used to measure the width

4
A.J. Lopes, et al. Additive Manufacturing 35 (2020) 101291

Fig. 1. Design of test part used for the ranking model and build orientation.
Adapted from Perez et al. [16].

Table 3 [65], an online service where private printer owners and FabLabs print
Design Features of Test Part in Fig. 1. parts for researchers and users at a low cost. The cost of the various
Letter Feature Factor Tested parts is also outlined in Table 4. Because the hubs were informed that
the part would be used for academic research purposes, many provided
A Square base Dimensional deviations the test part at a discounted price while others donated the test parts.
B (+) Lateral ridges
As a disclaimer, it is difficult to confirm that all of the instructions
C (–) Lateral ridges
D (+) Descending cylinders
were followed when requesting parts from such a large number of part
E (–) Descending cylinders providers. The authors had to trust that this was the case. Additionally,
F (+) Staircase the cost listed in Table 4 is the price to acquire the part from the various
G (–) Staircase sources and it was not possible to segment the cost into different ca-
H (+) Cylinders
tegories. Future studies, where the build may be better controlled by
I (–) Cylinders
J Ramps Surface roughness the researchers, will consider the total cost (including materials and
K Rectangular prisms Positional error post-processing) of making the part using the different technologies. It
L Tensile Bar Ultimate tensile strength should be noted that the actual cost of making the part may be different
M Thin angled wall Visual aesthetics from the price paid to acquire it, as listed in Table 4. For reference,
N (+) Dome
O (–) Dome
average costs for some of the materials used in this research are as
P Thin cylindrical wall follows (as per a 2015 search): PLA ($15-$20 per kg); ABS ($15-$20 per
Q (+) Descending rectangular prisms kg); and Nylon ($50-$73 per kg).
R (–) Descending rectangular prisms

3.3. Test part fabrication


including overhanging features in a test part [11–13]. However, it was
decided not to do so as overhanging features would have required ad- The goal of the project was to print one ABS or PLA test part on each
ditional support material. The need for removing support material from of the 45 systems as these were the two most commonly available
such features may introduce part variability since users may remove materials at the time of acquiring the test parts. However, some ex-
support differently, potentially affecting the measured features [13]. ceptions were made in the case of the Solido - SD300 Pro (PVC),
The design dimensions are provided in Fig. 2. Formlabs - Form1+ (resin), FSL3D - Pegasus Touch (resin), and
Markforged - Mark One (Nylon) as these systems cannot print ABS or
PLA. The two Stratasys systems use ABS, but it is sold as ABSplus which
3.2. Obtaining test parts has been tailored to work well with their FDM technology as well as
their support material. PLA has grown in popularity for its biocompa-
Seven of the 45 parts tested were printed by the authors using tible nature, which allows that it be used in medical applications [66].
systems housed within the W.M. Keck Center for 3D Innovation at the It is therefore, preferred by those who consider it more environmentally
University of Texas at El Paso (UTEP). To increase the sample size, the friendly than ABS. Stephens et al. determined that ultrafine particle
authors reached out to 24 manufacturers and universities, asking them emission rates in term of number of particles per minute ABS (1.9 ×
to print test parts using desktop systems available at their facilities. Of 1011 particles/min) was greater than of that PLA (2.0 × 1010 particles/
these, two manufacturers (Afinia and Stratasys) and four universities min) printed on desktop systems. They did not however, test the che-
(North Carolina State University, University of Akron, University of mical composition of these particles to determine toxicity levels [67]. It
Texas at Austin, and Virginia Tech) agreed to participate in the project was also asked of part providers that they only use raft support (i.e.
and donated test parts. The other parts were obtained through 3D Hubs support material that separates the model material from the build

5
A.J. Lopes, et al. Additive Manufacturing 35 (2020) 101291

Fig. 2. Design dimensions in mm, the letters given to each feature correspond to the letters in Fig. 1.
Adapted from Perez et al. [16].

platform) if any was needed. The part was designed so that additional density settings ranging from sparse-less dense (lowest density) to solid
support was not necessary, preventing damage to the parts during (highest density). The Markforged - Mark One, cannot print higher than
support removal. The SD300 Pro uses support regardless, due to the 50 % because the other 50 % consists of carbon fiber reinforcement. For
building nature of the sheet lamination process. Support was removed purposes of this study, the carbon fiber was not added. Finally, a layer
manually for all of the systems. It was also requested that the highest thickness of 100 μm was preferred, to include systems that cannot print
possible density setting be selected when building the part, preferably at higher resolutions. The first part successfully printed by each system
100 %. Some systems, like those from MakerBot, have percentage set- was used for purposes of this evaluation.
tings (1 %–100 %). Other systems, like those from Stratasys, have

6
A.J. Lopes, et al. Additive Manufacturing 35 (2020) 101291

Table 4 achieved by taking the measured value of a specific feature and sub-
Cost per Test Part (all costs at time of purchase, circa 2015). tracting it from the CAD value of that feature. A negative value in-
System Source of Part Cost Build Material dicated the fabricated feature was larger than the CAD, whereas a po-
sitive value indicated a smaller fabricated feature. For purposes of the
3D Systems - Cube U. of Akron D ABS ranking model, an absolute value was taken. A sum of all errors was
Afinia - H480 Afinia D ABS
calculated for each test part, providing one single value that was used in
Airwolf 3D - HD2X 3D Hubs $43.45 ABS
BeeVeryCreative - BeeTheFirst 3D Hubs $30.21 PLA
the ranking model. The optical system on the OGP Smartscope was used
Bits from Bytes - 3D Touch UTEP PIH ABS to measure the negative features (i.e. holes) on the test part. These too
Bits from Bytes - BFB 3000 NC State D ABS were compared to the CAD values. A value of 0 mm (and therefore
Bits from Bytes - RapMan 3.1 NC State D ABS relative error of 100 %) was given to features that the touch probe or
Builder 3D Printers - Builder Dual 3D Hubs D ABS
optical system could not detect because of their absence or lack of
Feed
Builder 3D Printers - Big Builder Dual 3D Hubs D PLA definition often caused by fabrication errors.
Feed
CEL - Robox 3D Hubs $16.81 ABS 3.5. Positional error
Craft Unique - CraftBot 3D Hubs $22.54 ABS
Dremel 3D - Idea Builder 3D Hubs $45.44 PLA
Felix - Felix 3.0 3D Hubs $38.44 PLA
Positional error (PE) along the X axis was calculated by measuring
Flashforge - Creator 3D Hubs $59.11 PLA the distance (d) from the first rectangular prism to the second rectan-
Flashforge - Dreamer 3D Hubs $13.12 ABS gular prism (Fig. 1 - feature K). The actual distance was subtracted from
Formlabs - Form1+ UTEP PIH Resin the CAD distance and the absolute value was taken. The same was done
FSL3D - Pegasus Touch 3D Hubs $40.27 Resin
for the other four rectangular prisms, always measuring the distance
Leapfrog - Creatr HS 3D Hubs $30.21 PLA
LulzBot - Mini 3D Hubs $51.63 ABS from the first rectangular prism. The sum of the errors was taken. The
LulzBot - TAZ 4 3D Hubs $42.71 ABS process was repeated for the rectangular prisms along the Y axis and
MakerBot - Replicator (1st Generation) UTEP PIH ABS both, the X and Y, values were used in the ranking model.
MakerBot Replicator (5th Gen.) UTEP PIH ABS
MakerBot - Replicator 2 UT Austin D PLA
MakerBot - Replicator 2X UTEP PIH ABS
3.6. Surface roughness
MakerBot - Replicator Mini 3D Hubs $30.21 PLA
MakerBot - Replicator Z18 3D Hubs $30.21 PLA To evaluate surface roughness, a Mitutoyo SJ-201 P surface
MakerBot - Thing-O-Matic Virginia Tech D ABS roughness tester (Mitutoyo America Corp., Aurora, IL) was used. Three
Markforged - Mark One 3D Hubs $47.05 Nylon
measurements were taken from each of the four inclined ramps. The
MBot - Cube 3D Hubs $33.40 PLA
PP3DP - Up! Plus UTEP PIH ABS mean Ra values of surface roughness for each ramp were added to give
Printrbot - Metal Plus 3D Hubs $23.58 PLA the test part one value for use in the ranking model. The ramps had
Printrbot - Simple Metal 3D Hubs $23.58 PLA various angles of inclination: 10°, 15°, 30°, and 45°. The measurements
Rapide 3D - Rapide Lite 200 3D Hubs $29.09 PLA were conducted from the bottom to the top of the ramp. Geometric
RepRap - Mendel 3D Hubs $51.63 ABS
RepRap - Mini Kossel 3D Hubs $51.63 ABS
models that can predict surface roughness and dimensional deviations
RepRap - Prusa Mendel U. of Akron D ABS while assessing the impacts of these deviations on part features, may
SeeMeCNC - Rostock Max 3D Hubs $51.63 ABS help optimize these features in AM parts [72,73].
Solido - SD300 Pro UTEP PIH PVC
Solidoodle - Solidoodle 4 3D Hubs $11.78 PLA
3.7. Mechanical testing
Stratasys - Mojo Stratasys D ABSplus
Stratasys - uPrint Plus UTEP PIH ABSplus
Type A Machines - Series 1 3D Hubs $55.44 PLA Mechanical testing was completed using one ASTM D638 Type V
Ultimaker - Ultimaker 2 3D Hubs $33.41 PLA specimen [74] built alongside the square base of the design in the XY
XYZ Printing - Da Vinci 2.0 Duo 3D Hubs $30.37 ABS orientation per ASTM F2971 (Fig. 1 - feature L). An Instron 5866 (In-
Zortrax - M200 3D Hubs $39.85 ABS
stron, Norwood, MA) tensile testing machine with a load cell of 10 kN, a
D: Part donated, PIH: Part printed in-house. ramp rate of 1 mm/min, and a load measurement accuracy of ± 0.4 %
(as per manufacturer’s specifications) was used to measure ultimate
3.4. Dimensional deviations tensile strength. As per ASTM D638 recommendations, the specimens
were preconditioned at 23 °C and 50 % relative humidity for 40 h.
Several researchers are utilizing a variety of artifact measuring
methods, including 3D scanning and X-ray computed tomography 3.8. Ranking model
(XCT) to gain valuable insights on, and effectively evaluate, the quality
of AM printers [68,69]. In 2019, McGregor et al. utilized a high-re- The systems were ranked based on nine equally weighted factors:
solution document scanner and XCT system to evaluate complex 2D and system build volume, system size, system weight, system cost, layer
3D lattice structures through statistical analysis of several key dimen- resolution, dimensional accuracy, surface roughness, and positional
sions [70]. In 2020, Santos et al., evaluated the contact coordinate error along the x and y-axes. These nine factors allowed for a consistent
measurement machine (CMM), a close-range photogrammetry (PG) method of comparison that removed biased evaluations such as the
system, and XCT system to highlight differences in measuring techni- user-friendliness of the system. Although the quantitative ranking
ques and the need for specification standards and guidance for such model described below was presented by Roberson et al. [15], the steps
assessments [71]. are reiterated here to give readers a comprehensive view of the ranking
For this research, dimensional deviations were evaluated using an process. Unlike the multiple iterations conducted in Roberson et al.,
OGP Smartscope Flash 250 (Optical Gaging Products, Rochester, NY) only one outlier removal step was conducted here. The multiple itera-
equipped with a TP200 modular probe (Renishaw Inc., Hoffman tion process was not suitable for the large number of systems tested
Estates, IL) due to accessibility. A series of points along the various here because the data were very widely spread and the iterations left
features of the 45 test parts were measured with a ruby ball styli (Ø = 1 very few systems in each category. The following steps were used to
mm). The obtained measurements were compared to those of the ori- determine the contribution of each factor to the total score of a system:
ginal CAD of the part and the relative error was computed. This was
1 The mean (x ) and standard deviations (s ) for each data set within a

7
A.J. Lopes, et al. Additive Manufacturing 35 (2020) 101291

factor were calculated in the following way: 3.9. Comparison with other commonly used models

xi
x = The ranking of the desktop 3D printers evaluated in this study was
n (1) compared to corresponding rankings provided by other commonly used
techniques for classifying entities using a multi-criteria decision making
n (x i x )2 method. The ranking models evaluated are discussed below:
s=
i=1 n 1 (2)
1 Simple Ranking Model:
2 The following formulas were used to calculate a threshold value ( ):
In this method, each 3D printing system was given a rank within
= x + s [Used when a low value of x was preferable, such as in SS, SC, each factor (1 (best) – 45 (worst); adjusted for ties). Then, all the ranks
MC, DD, Ra, XPE, and YPE] (3) across the nine factors were added to give an overall score for each
system. The 3D printer with the lowest total score was designated as the
= x s [Used when a high value of x was preferable, such as in BV] best system and the printer with the highest total score was ranked as
the worst system. Although adequate, this method does not provide the
(4)
magnitude at which a particular system is better or worse as compared
to the rest.
3 Each value of x was compared to the threshold (ϑ) to remove out-
liers if
2 Evaluating Percentile Data Rank (Factor Contribution, FC)
xi < [When a low value of x was preferable] (5)
For this method, the factor contribution (FC) values discussed in
xi > [When a high value of x was preferable] (6) step four of Section 3.8 were evaluated for each system, which provided
a percentile score of each system for each factor. The combined score of
These outliers were not taken as a measurement error, but rather a factor contribution values for each 3D system across all nine factors
measurement of the limitations of each system. The outlier removal were then taken into consideration to rank the systems from the best
process was a way of ensuring only competitive systems were rewarded (highest total FC score) to the worst (lowest total FC score). This
while others were removed from the respective factor. method does not consider any outliers within the comparison group for
each factor analyzed.
4 The remaining values were used to calculate each factor’s con-
tribution (FC). 3 Technique for Order Preference by Similarity to an Ideal Solution
(TOPSIS)
min (x )
FC = [When a low value of x was preferable]
xi (7)
The Technique for Order Preference by Similarity to the Ideal
xi Solution (TOPSIS) is aimed at evaluating, prioritizing, and selecting the
FC = [When a high value of x was preferable] best alternative that is closest to the ideal solution and farthest from the
max (x ) (8)
negative ideal alternative, based on calculated Euclidean distances. The
classical TOPSIS method utilizes information on all the key attributes in
5 The equally weighted contribution ( f (x ) ) was calculated for each the form of numerical data and the only subjective inputs are weights, if
system. applicable [75]. The key advantages of the TOPSIS method are in its
FC simplicity to order the data, rationality in its method, comprehensive-
f (x ) = ness in the steps, and its overall good computational efficiency to
FC (9)
measure the relative performance for each alternative in a simple
mathematical form. The TOPSIS method involves the following key
6 The weighted contribution ( f (x ) ) was scaled based on the number of
steps [75]:
systems that survived the outlier removal process. [Example: If only
40 of the 45 systems survived, the total number of points available
for each factor was 40/45 or 0.89. If all systems survived, the • Normalizing the data to have a consistent attribute-less value across
all factors
number of points available for each factor was 45/45 or 1.]
7 The scaled and weighted contributions (f (x )) for each system were • Determining positive and negative ideal solutions from normalized
data
summed and divided by the sum of f (x ) of all systems to determine
the ranking score (R). • Calculating the separation measures from the positive ideal solution
and the negative ideal solution

R=
sum of f (x ) for each system • Calculating closeness to the positive ideal solution – Performance
(10) Score
• Ranking in descending order of Performance Score
sum of f (x ) for all systems

The purpose of Step 5 was to ensure if all systems survived the


outlier removal process, the sum of all f(x) values would be equal to 1. The detailed equations for each step are provided in the article by E.
Step 6 guaranteed systems would not be excessively rewarded by Roszkowska [75].
having more points, if other systems were removed from a particular
factor. For purposes of this paper, the ranking model weighed each of 4. Results and discussion
the nine factors (build volume, system size, system weight, system cost,
layer resolution, dimensional deviations, surface roughness, and posi- 4.1. Dimensional accuracy
tional error along the x and y-axes) equally. However, the model may
be modified if the user decides to give more weight to one or more of The sum of dimensional deviations in each printed part is listed in
the factors. For this, each f(x) simply has to be multiplied by the Table 5. The best dimensional accuracy (i.e. lowest number of devia-
weighting fraction the user feels is appropriate. Another benefit is that tion) was seen in the Type A Machines - Series 1 (7.18 mm), Afinia -
it can be modified to include more factors or systems when desired. H480 (7.73 mm), and the RepRap - Prusa Mendel (8.28 mm) as shown

8
A.J. Lopes, et al. Additive Manufacturing 35 (2020) 101291

Table 5
Dimensional Deviations and Positional Error (PE) per Test Part.
System Sum of Dimensional Deviations (mm) PE along X-Axis (mm) PE along (mm) Ratio of PE (X/Y)

RESIN
Formlabs - Form1+ 11.57 2.8527 0.4550 6.2703
FSL3D - Pegasus Touch 58.57 8.3239 2.1613 3.8514
PVC
Solido - SD300 Pro 54.48 0.1200 0.1500 0.8000
NYLON
Markforged - Mark One 10.86 3.5992 1.6154 2.2280
PLA
BeeVeryCreative - BeeTheFirst 11.82 0.5234 0.5285 0.9903
Builder 3D Printers - Builder Dual Feed 37.45 5.6366 7.9276 0.7110
Dremel 3D - Idea Builder 19.60 1.7423 0.3380 5.1542
Felix - Felix 3.0 11.36 1.4583 1.1225 1.2991
Flashforge - Creator 15.45 0.4776 0.9754 0.4897
Leapfrog - Creatr HS 21.67 1.9917 2.1312 0.9345
MakerBot - Replicator 2 23.52 0.9985 0.4804 2.0787
MakerBot - Replicator Mini 19.67 0.8486 1.6056 0.5285
MakerBot - Replicator Z18 18.63 2.3092 1.9050 1.2121
MBot - Cube 19.06 0.9325 1.2472 0.7477
Printrbot - Metal Plus 33.35 1.9810 3.2946 0.6013
Printrbot - Simple Metal 20.72 5.6767 1.3715 4.1389
Rapide 3D - Rapide Lite 200 15.08 1.9453 6.8508 0.2840
Solidoodle - Solidoodle 4 41.50 1.4043 303,477 0.4195
Type A Machines - Series 1 7.18 2.4819 2.5043 0.9910
Ultimaker - Ultimaker 2 10.10 4.7984 1.9709 2.4347
ABS
3D Systems - Cube 57.98 5.5070 4.3310 1.2715
Afinia - H480 7.73 1.4119 0.2186 6.4584
Airwolf 3D - HD2X 21.06 0.9908 2.2253 0.4452
Bits from Bytes - 3D Touch 34.65 0.7400 0.9400 0.7872
Bits from Bytes - BFB 3000 51.12 0.9655 0.3076 3.1392
Bits from Bytes - RapMan 3.1 67.49 3.3300 3.2079 1.0381
Builder 3D Printers - Big Builder Dual Feed 37.02 1.3003 0.9628 1.3506
CEL - Robox 17.92 0.3426 2.0983 0.1633
Craft Unique - CraftBot 43.41 5.6110 2.6292 2.1341
Flashforge - Dreamer 18.42 0.8156 1.2728 0.6408
LulzBot - Mini 16.34 1.4805 3.7743 0.3923
LulzBot - TAZ 4 12.28 0.2791 0.3915 0.7130
MakerBot - Replicator (1st Generation) 37.80 4.3945 1.3741 3.1981
MakerBot Replicator (5th Gen.) 33.31 0.2593 1.4252 0.1820
MakerBot - Replicator 2X 17.31 2.5800 1.0400 2.4808
MakerBot - Thing-O-Matic 19.94 0.7342 2.6013 0.2823
PP3DP - Up! Plus 18.50 0.3960 1.2652 0.3130
RepRap - Mendel 32.30 1.7681 1.2672 1.3954
RepRap - Mini Kossel 22.76 6.6824 6.5885 1.0143
RepRap - Prusa Mendel 8.28 0.7846 0.2871 2.7329
SeeMeCNC - Rostock Max 19.79 1.1128 1.8824 0.5911
Stratasys - Mojo 13.18 5.1514 0.6324 8.1458
Stratasys - uPrint Plus 11.23 0.1000 0.1900 0.5263
XYZ Printing - Da Vinci 2.0 Duo 41.28 0.7037 1.6107 0.4369
Zortrax - M200 9.69 0.3883 1.4760 02,631

Fig. 3. Highest dimensional accuracy: (A) Type A Machines - Series 1, (B) Afinia - H480; (C) RepRap - Prusa Mendel.

in Fig. 3. This is an interesting result considering none of these systems FSL3D - Pegasus Touch (58.27 mm), and 3D Systems - Cube (57.98
have an enclosed build envelope. In addition, the only system with a mm). Of particular interest is the fact that while the H480, Up! Plus,
head build platform is the Afinia. On the other hand, the highest di- and Cube, all have similar system designs, the dimensional deviations
mensional deviations were seen with the RapMan 3.1 (67.49 mm), were very different. As previously mentioned, the features not

9
A.J. Lopes, et al. Additive Manufacturing 35 (2020) 101291

recognized by the touch probe or optical system of the OGP Smartscope 4.3. Surface roughness
Flash 250, were penalized with a dimensional value of 0. This specifi-
cally affected the test part printed on the SD300 Pro where many of the Fig. 4 illustrates the values of surface roughness obtained from each
features broke during removal of excess material. Unlike with material of four inclined planes (i.e. ramps). The mean Ra values acquired for
extrusion, the use of surrounding support is unavoidable in sheet la- each ramp were summed to obtain one total value for use in the ranking
mination 3D printing. model. The expectation was for a decrease in ramp angle to increase
surface roughness due to the staircase effect [64]. The Pearson corre-
lation between layer resolution and surface roughness was 0.56 and the
4.2. Positional error coefficient was statistically significant. The lowest degree of surface
roughness was seen in the Formlabs - Form 1+ (19.76 μm) with its
Table 5 also displays the results of positional error along the X and Y lowest values in the 45° plane. This is an expected result as the Form
axes. The original intent was to determine linear displacement error by 1+ is a vat photopolymerization technology, which has always been
taking the slope of the linear regression. However, the points did not known for its high quality surface finish. Other systems with low values
always behave in a linear fashion; in some cases yielding a parabolic were the FSL3D - Pegasus Touch (32.15 μm) and the Zortrax - M200
shape. Therefore, it was decided to instead take a positional error by (54.92 μm). On the opposite end of the spectrum, the highest value of
measuring each point’s error from its originally intended position. surface roughness, was actually seen with the Stratasys - uPrint Plus
These values were then added for all five points along the X-axis and (114.95 μm) with its highest values in the 30° plane. The MakerBot -
along the Y-axis. The purpose of capturing these errors was to capture Replicator 2 (114.76 μm) and the Bits from Bytes - RapMan 3.1 (113.24
the errors associated with XY motion and thermal shrinkage of each μm) also had very high degrees of surface roughness.
system. In a majority of cases (24 of 45 systems), positional error along Additional statistical analysis was performed to understand the ef-
the Y-axis was larger than along the X-axis. The lowest error in both fect of ramp angle on average surface roughness values, their corre-
instances was held by the Stratasys - uPrint Plus (X, 0.1000 mm and Y, sponding standard deviations, and coefficient of variation, across all
0.1900 mm). This is not surprising as the uPrint Plus has a temperature evaluated systems. Since the normality criteria was not satisfied for all
controlled build envelope and uses the matured FDM technology. Sli- three parameters (P values < 0.05), the non-parametric Kruskal-Wallis
cing operation and tooling paths have been well-developed by Stra- test was conducted to gain insights on effect of ramp angle on the three
tasys, incorporating this thermal shrinkage phenomenon. parameters. The values for surface roughness across all systems were
Despite operating via a laser process, the highest positional error statistically different for all four ramp angles. The standard deviation
along the X-axis was given by the FSL3D - Pegasus Touch (8.3239 mm). values for the four ramp angles were statistically similar (P value =
The highest positional error along the Y-axis was seen in the Builder 3D 0.289) across all 45 systems. Finally, analysis of the coefficient of
Printers - Big Builder Dual Feed (7.9276 mm). The ratio of positional variance (COV) indicated that the values for 15° and 45° ramps were
error along the X-axis over that along the Y-axis was also calculated to statistically similar and different from the 10° and 30° ramps. For future
determine which systems are affected by part warpage as well as dis- research, a statistical study comparing industrial AM systems, using a
play a large difference in build accuracy along the two axes. Systems potentially unified artifact design for evaluating surface roughness, will
with a ratio < 1 showed lower error along the X-axis, while systems be conducted.
with a ratio > 1 had greater error along the X-axis.

Fig. 4. Summarized surface roughness values categorized by material type.

10
A.J. Lopes, et al. Additive Manufacturing 35 (2020) 101291

Fig. 5. Mechanical testing results sorted by material type.

4.4. Mechanical testing results in the final ranking model.

Fig. 5 shows the results of ultimate tensile strength (UTS), a value


obtained from one tensile bar printed with each test part (Feature L). To 4.5. Ranking model
attempt the best possible mechanical properties, the research team
asked that each part provider print the parts as dense as possible. The The results of the ranking model are displayed in Fig. 6. Several
top UTS values were seen in PLA parts: the Flashforge - Creator (53.05 aspects contributed to these results. The highest overall scores were
MPa), the Felix 3.0 (51.86 MPa) and the Printrbot - Simple Metal (49.33 obtained by the Formlabs - Form 1+ (Fig. 7A) and the Afinia - H480
MPa). This was an expected behavior since the bulk UTS of PLA (Fig. 7B) which both received contributions in eight of nine factors. The
(110–145 MPa [76]) is greater than that of bulk ABS (37 MPa [77]). Form 1+ exceled in categories like layer resolution and surface
The highest UTS amongst ABS parts was in the Builder 3D Printers - roughness while the H480 had high scores in system size, system
Builder Dual Feed (45.16 MPa), RepRap - Mini Kossel (41.40 MPa), and weight, dimensional deviations, and positional error along the Y-axis.
LulzBot - Mini (40.36 MPa). The lowest UTS was seen in the two resin Despite only receiving four of nine scores, the uPrint Plus (Fig. 7C) did
parts built by the FSL3D - Pegasus Touch (11.54 MPa) and the Formlabs very well, placing 3rd. It will be shown in the next section that users not
- Form 1+ (15.64 MPa). The Markforged - Mark One only achieved a hindered by cost can rank the uPrint Plus higher. It specifically exceled
UTS value of 16.59 MPa because as previously mentioned, the Nylon in the two positional error categories. On the lower half of the graph are
used to build the test part did not include the carbon fiber reinforce- the Bits from Bytes - RapMan 3.1(Fig. 7D) and BFB 3000 (Fig. 7E), as
ment that this machine is popular for. Because of the broad range of well as the MakerBot - Replicator (1st Gen.) (Fig. 7F). These systems
materials tested in this category, it was decided not to include UTS received only five, seven, and seven of nine contributions, respectively.
One surprising system is the Stratasys - Mojo. While it built a high

11
A.J. Lopes, et al. Additive Manufacturing 35 (2020) 101291

Fig. 6. Ranking model results. Systems are sorted from highest ranking (top) to lowest ranking (bottom).

quality test part with low surface roughness and high dimensional ac- represents a ranking of the systems based solely on the system’s features
curacy, it was affected by its high system cost. The Mojo also has a small (factors included: build volume, system size, system cost, system
build envelope despite its fairly large system dimensions. weight, and layer resolution). This group was led the RepRap - Mini
Kossel, a system that is affordable and has a large build volume while
being lightweight. Some of the most expensive systems ranked toward
4.6. Other groupings
the bottom. Although it has a high cost, the MakerBot - Replicator Z18
ranked in the top 50 % for its large build volume. Group 4 only ana-
Table 6 represents other groupings created based on observed trends
lyzed the quality of the test part itself. This group was led by the uPrint
amongst the systems. The results described in the original ranking
Plus and the SD300 Pro. Unlike in the original ranking model, the Mojo
model (Fig. 6) are provided for comparison. The ranking model as de-
ranked much higher, moving from the 33rd spot to the 18th. It also hurt
scribed in Section 3.8 of this paper was used in all six groups. Group 1
some printers like the 3D Systems - Cube which went from the 21st
displays the rankings if mechanical testing values had been included.
position to the 42nd. It is important to note that three of the systems
This helped most of the systems with PLA test parts go up in rank. It also
ranked within the top ten positions in all four groups as well as in the
penalized the Formlabs - Form 1+, bringing it down to the third spot.
original ranking model; the Form 1+, the H480, and the BeeTheFirst.
In Group 2, system cost was removed from the ranking model. This
Group 5 represents the Stratasys-owned printers. The seven MakerBot
ranking may aid those who do not see high cost as an obstacle when
systems are included in this list as Stratasys acquired MakerBot. This
purchasing a new system, such as researchers with a stationary lab and
group was led by the Replicator Z18, one of the newest systems in the
substantial research funds. In this group, the uPrint Plus moved up to 1st
Replicator line (in 2015), and the uPrint Plus. The original MakerBot -
place as it was the most expensive system in the group. The Stratasys -
Replicator and the MakerBot - Thing-O-Matic, the two oldest systems of
Mojo also moved up substantially to the 23rd position. Group 3

12
A.J. Lopes, et al. Additive Manufacturing 35 (2020) 101291

Fig. 7. Highest (top row) and lowest (bottom) ranking test parts: (A) Formlabs - Form 1+, (B) Afinia - H480; (C) Stratasys - uPrint Plus, (D) MakerBot - Replicator (1st
Gen.), (E) Bits from Bytes - BFB 3000, and (F) Bits from Bytes - RapMan 3.1.

the group, ranked last. ideal solutions.


In addition to the correlation between layer resolution and surface
4.7. Comparison of ranking models roughness, the system size had a positive correlation with both build
volume and system weight. Consequently, further analysis on the data
Table 7 summarizes the results from all four ranking models. Upon for all raking models was conducted, while eliminating the layer re-
closer review of the rankings across all four ranking models, a few solution and system size factors from the analysis. Table 8 summarizes
exceptions were observed. The Stratasys - uPrint Plus and Solido - the rankings based on the remaining seven factors (build volume, unit
SD300 Pro, which ranked high using the model presented in this re- cost, system weight, dimensional deviation, surface roughness, X-posi-
search, were ranked lower in the other rankings. It was observed that tional error, and Y-positional error). It was observed that across all
although these two systems ranked high in some factors, they also ranking models the following systems present the top three desktop 3D
ranked near the bottom for many other factors. There were several printer alternatives to the user: LulzBot TAZ 4, Rep-Rap Prusa Mendel
outlier systems for the two factors (X and Y Positional Error) in which (1st Generation), and Dremel 3D Idea Builder.
these two systems were at the top of the list. As a result, their high score Finally, statistical analysis was performed on the data for the seven
within those two factors placed them at a high ranking in the original non-correlated factors to determine effect of each manufacturer set
ranking model. The other models penalized their low rankings in other parameters (build volume, unit cost, and system weight) on the four
factors more than the ranking model presented in this paper, and as a measured performance parameters (dimensional deviation, surface
result resulted in lower rankings. roughness, X-positional error, and Y-positional error). To facilitate the
Additionally, it was observed that three systems - Builder 3D analysis, the manufacturer set parameters were categorized as shown in
Printers-Big Builder Dual Feed, SeeMeCNC-Rostock Max, and Airwolf- Table 9 below:
HD2X ranked higher in the TOPSIS rankings as compared to the other The Kruskal–Wallis test is a non-parametric method that does not
three ranking methods. This simply indicates that across nine, equally assume that the response variable is normally distributed, which was
weighted factors, these three systems were closer to the ideal solution the case with the data analyzed for this research. It was observed that
based on its relative Euclidean distances from the positive and negative there was no statistical differences (α = 0.05) between the measured

13
Table 6
Other Groupings.
Group 1 Group 2 Group 3 Group 4 Group 5
A.J. Lopes, et al.

Original Ranking Model Results (top rank Results Including Mechanical Testing Cost Not an Issue Analyzing System Features Analyzing Test Part Quality Stratasys - Owned Systems
first)

Formlabs - Form 1+ Afinia - H480 Stratasys- uPrint Plus RepRap - Mini Kossel Stratasys- uPrint Plus MakerBot - Replicator Z18
Afinia - H480 Stratasys- uPrint Plus Formlabs - Form 1+ CEL - Robox Solido - SD300 Pro Stratasys- uPrint Plus
Stratasys- uPrint Plus Formlabs - Form 1+ Afinia - H480 3D Systems - Cube (1st Gen.) Formlabs - Form 1+ MakerBot - Replicator Mini
Solido - SD300 Pro Solido - SD300 Pro Solido - SD300 Pro Ultimaker - Ultimaker 2 Afinia - H480 Stratasys - Mojo
LulzBot - TAZ 4 LulzBot - TAZ 4 LulzBot - TAZ 4 PP3DP - Up! Plus (1st Gen.) LulzBot - TAZ 4 MakerBot - Replicator (5th
Gen.)
CEL - Robox CEL - Robox CEL - Robox Builder 3D Printers - Big Builder Dual RepRap - Prusa Mendel (1st Gen.) MakerBot - Replicator 2X
Feed
st st
RepRap - Prusa Mendel (1 Gen.) RepRap - Prusa Mendel (1 Gen.) BeeVeryCreative - BeeTheFirst Afinia - H480 BeeVeryCreative - BeeTheFirst MakerBot - Replicator 2
BeeVeryCreative - BeeTheFirst RepRap - Mini Kossel Zortrax - M200 BeeVeryCreative - BeeTheFirst Zortrax - M200 MakerBot - Replicator (1st
Gen.)
RepRap - Mini Kossel BeeVeryCreative - BeeTheFirst Type A Machines- Series 1 Formlabs - Form 1+ Dremel 3D - Idea Builder MakerBot - Thing-O-Matic
Zortrax - M200 Zortrax - M200 Ultimaker - Ultimaker 2 Type A Machines- Series 1 MakerBot - Replicator (5th Gen.)
PP3DP - Up! Plus (1st Gen.) Type A Machines- Series 1 PP3DP - Up! Plus (1st Gen.) Rapide 3D - Rapide Lite 200 CEL - Robox
Ultimaker - Ultimaker 2 PP3DP - Up! Plus (1st Gen.) RepRap - Prusa Mendel (1st Gen.) Printrbot - Simple Metal Flashforge - Dreamer
Type A Machines- Series 1 Felix - Felix 3.0 RepRap - Mini Kossel RepRap - Mendel Type A Machines- Series 1
Dremel 3D - Idea Builder Dremel 3D - Idea Builder Builder 3D Printers - Big Builder Dual Zortrax - M200 PP3DP - Up! Plus (1st Gen.)
Feed
Felix - Felix 3.0 Ultimaker - Ultimaker 2 Felix - Felix 3.0 Solidoodle - Solidoodle 4 Felix - Felix 3.0
Builder 3D Printers - Big Builder Dual Builder 3D Printers - Big Builder Dual Dremel 3D - Idea Builder FSL3D - Pegasus Touch Flashforge - Creator
Feed Feed
Flashforge - Dreamer Flashforge - Creator MakerBot - Replicator Z18 MakerBot - Replicator Z18 Bits from Bytes - 3D Touch

14
FSL3D - Pegasus Touch Printrbot - Simple Metal FSL3D - Pegasus Touch Leapfrog - Creatr HS Stratasys - Mojo
Printrbot - Simple Metal Mbot - Cube Airwolf - HD2X Felix - Felix 3.0 Markforged - Mark One
Rapide 3D - Rapide Lite 200 MakerBot - Replicator Z18 Leapfrog - Creatr HS SeeMeCNC - Rostock Max Airwolf - HD2X
3D Systems - Cube (1st Gen.) Flashforge - Dreamer Flashforge - Dreamer RepRap - Prusa Mendel (1st Gen.) MakerBot - Replicator 2
Flashforge - Creator Rapide 3D - Rapide Lite 200 MakerBot - Replicator (5th Gen.) Mbot - Cube Ultimaker - Ultimaker 2
Mbot - Cube Leapfrog - Creatr HS Stratasys - Mojo LulzBot - Mini Mbot - Cube
Leapfrog - Creatr HS Airwolf - HD2X Rapide 3D - Rapide Lite 200 Printrbot - Metal Plus MakerBot - Replicator Mini
MakerBot - Replicator Z18 Printrbot - Metal Plus Flashforge - Creator Craft Unique - CraftBot MakerBot - Replicator 2X
Airwolf - HD2X MakerBot - Replicator (5th Gen.) Mbot - Cube LulzBot - TAZ 4 Builder 3D Printers - Big Builder Dual
Feed
MakerBot - Replicator (5th Gen.) LulzBot - Mini LulzBot - Mini Dremel 3D - Idea Builder FSL3D - Pegasus Touch
Solidoodle - Solidoodle 4 SeeMeCNC - Rostock Max Bits from Bytes - 3D Touch Airwolf - HD2X RepRap - Mini Kossel
LulzBot - Mini FSL3D - Pegasus Touch 3D Systems - Cube (1st Gen.) Builder 3D Printers -Builder Dual Feed Leapfrog - Creatr HS
SeeMeCNC - Rostock Max Solidoodle - Solidoodle 4 MakerBot - Replicator 2 Flashforge - Creator Printrbot - Simple Metal
Printrbot - Metal Plus MakerBot - Replicator 2 MakerBot - Replicator Mini XYZprinting - Da Vinci 2.0 Duo MakerBot - Replicator Z18
MakerBot - Replicator Mini 3D Systems - Cube (1st Gen.) Markforged - Mark One Flashforge - Dreamer Rapide 3D - Rapide Lite 200
Stratasys - Mojo MakerBot - Replicator Mini Printrbot - Metal Plus MakerBot - Replicator Mini XYZprinting - Da Vinci 2.0 Duo
MakerBot - Replicator 2 Stratasys - Mojo Printrbot - Simple Metal MakerBot - Thing-O-Matic LulzBot - Mini
XYZprinting - Da Vinci 2.0 Duo XYZprinting - Da Vinci 2.0 Duo SeeMeCNC - Rostock Max MakerBot - Replicator 2 Printrbot - Metal Plus
Bits from Bytes - 3D Touch Builder 3D Printers -Builder Dual Feed MakerBot - Replicator 2X Stratasys - Mojo MakerBot - Thing-O-Matic
Craft Unique - CraftBot RepRap - Mendel Builder 3D Printers -Builder Dual Feed MakerBot - Replicator (5th Gen.) SeeMeCNC - Rostock Max
RepRap - Mendel Craft Unique - CraftBot Solidoodle - Solidoodle 4 Bits from Bytes - RapMan 3.1 MakerBot - Replicator (1st Gen.)
Markforged - Mark One MakerBot - Thing-O-Matic MakerBot - Thing-O-Matic MakerBot - Replicator 2X Builder 3D Printers -Builder Dual Feed
Builder 3D Printers -Builder Dual Feed MakerBot - Replicator 2X Craft Unique - CraftBot Bits from Bytes - BFB 3000 Solidoodle - Solidoodle 4
MakerBot - Replicator 2X Bits from Bytes - 3D Touch XYZprinting - Da Vinci 2.0 Duo Bits from Bytes - 3D Touch Craft Unique - CraftBot
MakerBot - Thing-O-Matic Markforged - Mark One MakerBot - Replicator (1st Gen.) MakerBot - Replicator (1st Gen.) 3D Systems - Cube (1st Gen.)
MakerBot - Replicator (1st Gen.) MakerBot - Replicator (1st Gen.) RepRap - Mendel Markforged - Mark One Bits from Bytes - BFB 3000
Bits from Bytes - BFB 3000 Bits from Bytes - BFB 3000 Bits from Bytes - BFB 3000 Solido - SD300 Pro RepRap - Mendel
Bits from Bytes - RapMan 3.1 Bits from Bytes - RapMan 3.1 Bits from Bytes - RapMan 3.1 Stratasys- uPrint Plus Bits from Bytes - RapMan 3.1
Additive Manufacturing 35 (2020) 101291
A.J. Lopes, et al. Additive Manufacturing 35 (2020) 101291

Table 7
Comparison of Ranking Models – ALL Nine Factors Considered.

15
A.J. Lopes, et al. Additive Manufacturing 35 (2020) 101291

Table 8
Comparison of All Ranking Models – Only Seven Non-Correlated Factors Considered.

16
A.J. Lopes, et al. Additive Manufacturing 35 (2020) 101291

Table 9 Declaration of Competing Interest


Parameter Categorization for Statistical Analysis.
Parameter Low Medium High One or more of the authors of this article are a part of the Editorial
3 3
Board of the journal. To avoid a potential conflict of interest, the re-
Build Volume < 4000 cm > = 4000 cm and > = 10,000
sponsibility for the editorial and peer-review process of this article lies
< 10,000 cm3 cm3
Unit Cost < $2000 > = $2000 and < $4000 > = $4000 with a different editor. Furthermore, the authors of this article were
System Weight < = 10 kg > 10 kg and < 20 kg > 20 kg removed from the peer review process and had no, and will not have,
any access to confidential information related to the editorial process of
this article.
parameters when compared across all levels of the three manufacturer
set parameters. This illustrates that differences in build volume, unit Acknowledgements
cost, and system weight did not have any statistical impact on the
measured parameters, when compared using the above-categorized le- The research described here was performed at The University of
vels. In summary, percentile value based rankings models (which only Texas at El Paso (UTEP) within the W.M. Keck Center for 3D Innovation
consider how an alternative compares to the best solution) such as the (Keck Center). This research was supported by funding from the UTEP
one presented in this research, provide vastly different results com- Louis Stokes Alliance for Minority Participation program funded
pared to traditional separation based multi-criteria decision making through National Science Foundation grant number HRD-1202008, the
models such as TOPSIS (which evaluate each solutions’ separation from Texas Emerging Technology Fund, the Mr. and Mrs. MacIntosh
both ideal and worst solutions). Murchison Chair I in Engineering Endowment, and strategic invest-
ments in this research via discretionary UTEP Keck Center funds. The
authors are grateful to Nikki Martinez, Mohammad S. Hossain, Alfonso
5. Conclusions Fernandez, Alejandro Cuaron, Nate Johnston, Juan Vargas, and Elaine
MacDonald for their support in various aspects of this project. In par-
The current market demand for new and innovative desktop 3D ticular, the authors are grateful to Lluvia Herrera for performing au-
printers has resulted in an influx of new printer designs. Small afford- tomatic touch probe measurements and mechanical testing, and
able 3D printing systems are available for both the at-home user and grateful to Jorge Ramos for his effort in designing the test artifact used
engineer/professional seeking to find rapid design verifications. The in this study. The authors would also like to acknowledge Afinia,
work presented here focused on using a ranking model with 45 (circa Stratasys, North Carolina State University, the University of Akron, the
2015) desktop 3D printers based on factors a consumer may consider University of Texas, and Virginia Tech University for providing test
during their purchase. To evaluate the ranking model, a test part was parts included in the ranking model. Finally, the authors wish to thank
designed that included various measureable geometric features to test 3D Hubs for their help when acquiring test parts through their online
each system’s build accuracy. The ranking model found the Formlabs - service.
Form 1+ and Afinia - H480 to be the best systems due, in part, to the
Form1+’s outstanding surface finish and the dimensional accuracy of Appendix A. Supplementary data
the H480. In fact, their high rank persuaded the authors to purchase
both systems as they did not previously own them. Another system Supplementary material related to this article can be found, in the
added to the Keck Center repertoire is the LulzBot TAZ 5, and more online version, at doi:https://doi.org/10.1016/j.addma.2020.101291.
recently, TAZ 6, as a result of the TAZ 4′s excellent performance. When
comparing the ranking model discussed in this paper with other References
common ranking methods, it was observed that there were a few sys-
tems that ranked differently across the analyzed models. The resulting [1] "The Journey of a Lifetime," [Online]. Available: http://www.3dsystems.com/30-
analysis indicated that those systems that did exceedingly well in fac- years-innovation.
tors which eliminated many other systems as outliers, ranked very high [2] T.T. Wohlers, Wohlers Report 2018: 3D Printing and Additive Manufacturing State
of the Industry," Annual Worldwide Progress Report, Wohlers Associates Inc., Fort
in the model. Consequently, future research will focus on identifying Collins, CO, 2018.
the selection of ideal factors to be evaluated depending on the most [3] T.T. Wohlers, Wohlers Report 2015: 3D Printing and Additive Manufacturing State
appropriate application for desktop 3D printers and updating the model of the Industry, Wohlers Associates Inc., Fort Collins, CO, 2015.
[4] M. Richardson, B. Haylock, Designer/maker: the rise of additive manufacturing,
to eliminate this bias. Nonetheless, the LulzBot TAZ 4, RepRap Prusa domestic-style production and the possible implications for the automotive in-
Mendel (1st Gen.), and Dremel 3D Idea Builder, ranked high across all dustry, Comput. Aided Design Appl., PACE 2 (2012) 33–48.
four rankings models and could be recommended highly, at the time of [5] R. Jones, P. Haufe, E. Sells, P. Iravani, V. Olliver, C. Palmer, A. Bowyer, RepRap -
the replicating rapid prototyper, Robotica 29 (January) (2011) 177–191.
this analysis, to potential users. This research demonstrated that the [6] J.M. Pearce, C. Morris Blair, K.J. Laciak, R. Andrews, A. Nosrat, I. Zelenika-Zovko,
selected evaluation factors and type of ranking method have a clear 3-D printing of open source appropriate technologies for self-directed sustainable
impact on how the systems are ranked. The authors believe models development, J. Sustain. Dev. 3 (4) (2010) 17–29.
[7] F.A. Cruz Sanchez, H. Boudaoud, L. Muller, M. Camargo, Towards a standard ex-
similar to the one described here provide tremendous utility in the
perimental protocol for open source additive manufacturing, Virtual Phys. Prototyp.
evaluation of 3D printers and will continue to explore improvements. 9 (3) (2014) 151–167.
Future work will evaluate new and updated 3D desktop printers as well [8] ISO/ASTM 52900:2015: Terminology for AM – General Principles – Terminology,
as professional-grade systems, using optimally selected appropriate ASTM International, West Conshohocken, PA, 2015.
[9] T.H.C. Childs, N.P. Juster, Linear and geometric accuracies from layer manu-
factors evaluated with an updated ranking model, guided by the lessons facturing, CIRP Ann. Manuf. Technol. 43 (1) (1994) 163–166.
learned while comparing the list from the four ranking models eval- [10] J.G. Zhou, D. Herscovici, C.C. Chen, Parametric process optimization to improve the
uated as presented in this paper. The ability to build spatially variant accuracy of rapid prototyped stereolithography parts, Int. J. Mach. Tools Manuf. 40
(3) (2000) 363–379.
parts and print parts in different colors as well as other new features [11] M. Mahesh, Y.S. Wong, J.Y. Fuh, H.T. Loh, Benchmarking for comparative eva-
that enter the market will also be included as factors within an updated luation of RP systems and processes, Rapid Prototyp. J. 10 (2) (2004) 123–135.
ranking model. As these ranking models and measurement/inspection [12] Espalin D., Medina F., Arcaute K., Zinniel B., Hoppe T., Wicker R. B., "Effects of
vapor smoothing on ABS part dimensions," Proceedings from Rapid 2009
methods improve, more data should become available, leading to the Conference & Exposition, Schaumburg, IL.
realization of a consumer-specific tunable ranking model that can ra- [13] S. Moylan, J. Slotwinski, A. Cooke, K. Jurrens, M.A. Donmez, Proposal for stan-
pidly evaluate and rank industrial and desktop 3D printing systems dardized test artifact for additive manufacturing machines and processes,
Proceedings of the 2012 Annual International Solid Freeform Fabrication
based on specific consumer preferences.

17
A.J. Lopes, et al. Additive Manufacturing 35 (2020) 101291

Symposium, August, 2012, pp. 6–8. [48] UP! User Manual, PP3DP, (2012) [Online]. Available: http://www.pp3dp.com/
[14] Make: Annual Guide to 3D Printing, Maker Media, Inc., Sebastopol, CA, 2014, pp. index.php?option=com_jdownloads&Itemid=92&view=finish&cid=36&
30–71. catid=3.
[15] D.A. Roberson, D. Espalin, R.B. Wicker, 3D printer selection: a decision-making [49] Plus (10x10x10), printrbot, 2015 [Online]. Available: http://printrbot.com/
evaluation and ranking model, Virtual Phys. Prototyp. 8 (3) (2013) 201–212. product-category/3d-printers/metal-plus/.
[16] M. Perez, J. Ramos, D. Espalin, M. Hossain, R. Wicker, Ranking models for 3D [50] Simple (6x6x6), printrbot, 2015 [Online]. Available: http://printrbot.com/product-
printers, Proceedings of the 2013 Annual International Solid Freeform Fabrication category/3d-printers/simple-metal/.
Symposium (2013). [51] Rapide Lite, Affordable High Resolution 3D Printer, Indiegogo, 2015 [Online].
[17] Swanson W.J., Turley P.W., Leavitt P. J., Karwoski P. J., LaBossiere E., Skubic R. L., Available: https://www.indiegogo.com/projects/rapide-lite-affordable-high-
"High Temperature Modeling Apparatus". United States Patent US 6,722,872 B1, resolution-3d-printer#/.
2004. [52] Mendel, RepRap, 2015 [Online]. Available: http://reprap.org/wiki/Mendel.
[18] Q. Sun, G.M. Rizvi, C.T. Bellehumeur, P. Gu, Effect of processing conditions on the [53] Kossel, RepRap, 2015 [Online]. Available: http://reprap.org/wiki/Kossel.
bonding quality of FDM polymer filaments, Rapid Prototyp. J. 14 (2008) 72–80. [54] Prusa Mendel Build Manual, (2012) [Online]. Available: http://www.
[20] Cube, 3D Systems, (2014) [Online]. Available: http://cubify.com/en/Products/ nextdayreprap.co.uk/prusa-mendel-build-manual/.
Cube2TechSpecs. [55] Rostock Max, RepRap, 2014 [Online]. Available: http://reprap.org/wiki/Rostock_
[21] Afinia H480 3D Printer, Afinia 3D, (2014) [Online]. Available: http://www.afinia. max.
com/3d-printers. [56] Solido SD300 Pro, Solido Ltd., 2009 [Online]. Available: http://www.solido3d.
[22] AW3D HD2X, (2015) [Online]. Available: http://airwolf3d.com/shop/large-3d- com/default.htm.
printer-for-sale-now-aw3d-hd2x. [57] Solidoodle 4. Solidoodle, 2015 [Online]. Available: http://www.solidoodle4.com/
[23] BeeTheFirst, (2014) [Online]. Available: https://beeverycreative.com/beethefirst/. Solidoodle4.
[24] "3D Touch," Bits from Bytes, [Online]. Available: http://www.bitsfrombytes.com/ [58] Mojo, Stratasys Ltd., 2014 [Online]. Available: http://www.stratasys.com/3d-
sites/www.bitsfrombytes.com/files/0911-TOUCH-UK-3DS.pdf. printers/idea-series/mojo.
[25] " BFB-3000," Bits from Bytes, [Online]. Available: http://cubify.s3.amazonaws. [59] uPrint SE Plus, Stratasys Ltd., 2013 [Online]. Available: http://www.stratasys.com/
com/public/bfb/bfb_3000_specs_a4.pdf. 3d-printers/idea-series/uprint-se-plus.
[26] " RapMan 3.1," Bits from Bytes, [Online]. Available: http://www.bitsfrombytes. [60] Series 1., Type A Machines, (2015) [Online]. Available: http://www.
com/sites/www.bitsfrombytes.com/files/downloads/RapMan31_Specs_A4.pdf. typeamachines.com/series-1.
[27] "Builder – Dual-Feed Extruder," 2014 - 2015. [Online]. Available: http:// [61] Ultimaker 2. Ultimaker, 2015 [Online]. Available: https://ultimaker2.com/en/
3dprinter4u.nl/shop/3d-printers/3d-printer-builder-dual-feed-extruder/. products/ultimaker-2.
[28] "Big Builder – Dual-Feed Extruder," 2014 - 2015. [Online]. Available: http:// [62] Da Vinci 2.0 Duo, XYZ Printing, 2014 [Online]. Available: http://us.xyzprinting.
3dprinter4u.nl/shop/3d-printers/big-builder-dual-feed-extruder/. com/us_en/Product/da-Vinci-2.0?gclid=CLzv8Ky5_MgCFQumaQodhnkOOA.
[29] Robox, (2015) [Online]. Available: http://www.cel-robox.com/. [63] Zortrax M200, Zortrax, 2015 [Online]. Available: https://zortrax.com/printers/
[30] "CraftBot", 2008 - 2014. [Online]. Available: https://craftunique.com/item/ zortrax-m200/.
craftbot-anthracite-gray. [64] P.M. Pandey, N.V. Reddy, S.G. Dhande, Improvement of surface finish by staircase
[31] " Dremel 3D Idea Builder," [Online]. Available: https://3dprinter.dremel.com/3d- machining in fused deposition modeling, J. Mater. Process. Technol. 132 (1–3)
printer-benefits. (2003) 323–331.
[32] 3. Natives, "3D Natives," [Online]. Available: https://www.3dnatives.com/en/3D- [65] "3D Hubs," [Online]. Available: https://www.3dhubs.com/.
compare/imprimante/felix-3-0-kit. [Accessed 22 January 2020]. [66] M.L. Hans, A.M. Lowman, Biodegradable nanoparticles for drug delivery and tar-
[33] Creator, (2015) [Online]. Available: http://www.flashforge-usa.com/creator-3d- geting, Curr. Opin. Solid State Mater. Sci. 6 (4) (2002) 319–327.
printer/. [67] B. Stephens, P. Azimi, Z. El Orch, T. Ramos, Ultrafine particle emissions from
[34] "Dreamer," [Online]. Available: http://www.ff3dp.com/index.php/3d-printers/ desktop 3D printers, Atmos. Environ. 79 (2013) 334–339.
dreamer-3d-printer.html. [68] A. Karme, A. Kallonen, V.-P. Matilainen, H. Piili, A. Salminen, Possibilities of CT
[35] Form 1+ High-Resolution 3D Printer, Formlabs, 2015 [Online]. Available: http:// scanning as analysis method in laser additive manufacturing, Phys. Procedia 78
formlabs.com/products/3d-printers/form-1-plus/. (2015) 347–356.
[36] Pegasus Touch, FSL4D, (2015) [Online]. Available: http://www.fsl3d.com/ [69] A. Heralić, A.-K. Christiansson, B. Lennartson, Height control of laser metal-wire
pegasus-touch/. deposition based on iterative learning control and 3D scanning, Opt. Lasers Eng. 50
[37] "Creatr H.S.," Leapfrog 3D Printers, [Online]. Available: http://www.lpfrg.com/en/ (9) (2012) 1230–1241.
creatr-hs. [70] D.J. McGregor, S. Tawfick, W.P. King, Automated metrology and geometric analysis
[38] "LulzBot Mini 3D Printer," LulzBot, [Online]. Available: https://www.lulzbot.com/ of additively manufactured lattice structures, Addit. Manuf. 28 (2019) 535–545.
products/lulzbot-mini-3d-printer. [71] V.M. Rivas Santos, A. Thompson, D. Sims-Waterhouse, I. Maskery, P. Woolliams,
[39] "LulzBot TAZ Desktop 3D Printer," LulzBot, [Online]. Available: http://download. R. Leach, Design and characterisation of an additive manufacturing benchmarking
lulzbot.com/TAZ/documentation/current/LulzBot_TAZ-Brochure.pdf. artefact following a design-for-metrology approach, Addit. Manuf. 32 (2020).
[40] "MakerBot Replicator," MakerBot, 2009-2015. [Online]. Available: http://store. [72] J.-Y. Dantan, Z. Huang, E. Goka, L. Homri, A. Etienne, N. Bonnet, M. Rivette,
makerbot.com/replicator. Geometrical variations management for additive manufactured product, CIRP Ann.
[41] MakerBot Replicator 2, MakerBot, 2014 [Online]. Available: http://store.makerbot. Manuf. Technol. 66 (1) (2017) 161–164.
com/replicator2. [73] Z. Zhu, N. Anwer, Q. Huang, L. Mathieu, Machine learning in tolerancing for ad-
[42] "MakerBot Replicator 2X", MakerBot, 2009-2013. [Online]. Available: http://store. ditive manufacturing, CIRP Ann. Manuf. Technol. 67 (1) (2018) 157–160.
makerbot.com/replicator2x. [74] ASTM Standard D638-10: Standard Test Method for Tensile Properties of Plastics,
[43] "MakerBot Replicator Mini," MakerBot, 2009-2015. [Online]. Available: http:// ASTM International, 2010.
store.makerbot.com/replicator-mini. [75] E. Roszkowska, Multi-criteria decision making models by applying the TOPSIS
[44] "MakerBot Replicator Z18," MakerBot, 2009-2015. [Online]. Available: http:// method to crisp and interval data, Multiple Criteria Decision Making, Katowice,
store.makerbot.com/replicator-z18. 2011, pp. 200–230.
[45] MakerBot Thing-O-Matic, MakerBot, 2014 [Online]. Available: http://www. [76] "Ingeo™ Biopolymer 4043D Technical Data Sheet," [Online]. Available: http://
makerbot.com/support/thingomatic/troubleshooting/. www.natureworksllc.com/∼/media/technical_resources/technical_data_sheets/
[46] The Mark One, Markforged, 2015 [Online]. Available: https://markforged.com/ technicaldatasheet_4043d_films_pdf.pdf.
mark-one/. [77] ABSplus-P430, (2014) [Online]. Available: http://www.stratasys.com/∼/media/
[47] MBot Cube Desktop 3D Printer, MBot3D, (2014) [Online]. Available: http://www. Main/Secure/Material%20Specs%20MS/Fortus-Material-Specs/Fortus-MS-
mbot3d.com/product-3d-printer/mbot-cube. ABSplus-01-13-web.pdf..

18

You might also like