You are on page 1of 12

Additive Manufacturing 56 (2022) 102904

Contents lists available at ScienceDirect

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

Research paper

Reducing production losses in additive manufacturing using overall


equipment effectiveness
Shreeja Basak a, *, Martin Baumers a, Matthias Holweg b, Richard Hague a, Chris Tuck a
a
Centre for Additive Manufacturing, Faculty of Engineering, University of Nottingham, Nottingham NG8 1BB, United Kingdom
b
Saïd Business School, University of Oxford, Park End Street, Oxford OX1 1HP, United Kingdom

A R T I C L E I N F O A B S T R A C T

Keywords: Additive manufacturing (AM) has widely demonstrated its ability to economically produce parts at low volumes.
Overall equipment effectiveness Attention is now shifting to higher volume applications, such as mass customization and the manufacture of
Additive manufacturing standard parts. In these contexts, production losses due to process variability and inefficient machine use are
Process capability
common concerns for the application of AM. The Overall Equipment Effectiveness (OEE) metric is a tool widely
Capacity utilisation
used in traditional manufacturing to assess the effective use of production capacity. Previous studies seeking to
Variability
Efficiency apply OEE to AM are limited in scope and have neglected the above sources of inefficiency. This article seeks to
address this gap in two ways. First, we present a framework for measuring OEE within AM operations; and
systematically map the ‘six production losses’ to the AM workflow to codify our understanding of the main
sources of inefficiency. Second, we conduct a simulation study investigating how the AM operations approach,
product variety and lead time requirements affect the OEE of an AM process. Our findings demonstrate that –
with some conceptual adaptation – the OEE metric can indeed be used in the context of AM. Furthermore, we
identify the approach to AM operation as a major determinant of performance in terms of the OEE achieved. We
conclude with a set of managerial insights on how to apply the OEE metric to AM processes in practice.

1. Introduction AM in high-volume manufacturing, such as poor reliability and slow


process speed [9]. Nevertheless, recognising the importance of the
Additive manufacturing (AM) is gaining traction as a direct digital aforementioned advantages, efforts are being made to improve the
manufacturing method across industries, from medical devices to applicability of AM for mainstream manufacturing via increased rates of
aerospace and automotive [1]. Successful applications of AM harness its material deposition [10], and exploiting hardware, software and man­
geometric freedom, granted by the layer-by-layer operating principle, to agement to reduce process uncertainty [11]. In this higher volume
better meet the needs of the end-user via customization or improved manufacturing context, appropriate manufacturing performance mea­
functionality. Examples include unique hearing aids and dental implants surement is critical for tracking the current operations state of the
[2], custom shoe soles [3], and lightweight, fuel-efficient components in manufacturing system, monitoring the effect of improvement initiatives
aircraft [4]. Moreover, the tool-less nature of AM can simplify supply and evaluating manufacturing management decisions [12,13].
chains and enable more responsive, sustainable production, with bene­ The Overall Equipment Effectiveness (OEE) is a classic measurement
fits shown for automotive and machine tooling industries [5], as well as metric used in manufacturing, where it used as a comprehensive indi­
on-demand military spare parts production [6]. cator for effective capacity utilisation [13]. It covers availability, per­
While AM performance is well-established in lower volume appli­ formance, and quality and thus is often tracked over time in the context
cations, the above examples indicate benefits that become accessible on of continuous improvement initiatives [14]. It can also direct operations
a larger scale as AM enters mainstream manufacturing. As the technol­ managers towards sources of production losses and hidden capacity
ogy and business models both evolve, AM has the potential to disrupt within the system [15]. OEE is widely adopted in industry and partic­
mainstream traditional manufacturing, particularly for complex product ularly effective where the cost of lost capacity is high [16], which is
applications [7,8]. However, operational issues become a concern for relevant for AM as indicated by AM cost models [9].

* Corresponding author.
E-mail addresses: shreeja.basak@nottingham.ac.uk (S. Basak), martin.baumers@nottingham.ac.uk (M. Baumers), matthias.holweg@sbs.ox.ac.uk (M. Holweg),
richard.hague@nottingham.ac.uk (R. Hague), christopher.tuck@nottingham.ac.uk (C. Tuck).

https://doi.org/10.1016/j.addma.2022.102904
Received 18 January 2022; Received in revised form 9 May 2022; Accepted 16 May 2022
Available online 19 May 2022
2214-8604/© 2022 The Author(s). Published by Elsevier B.V. This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/).
S. Basak et al. Additive Manufacturing 56 (2022) 102904

Prior work in applying OEE to AM has shown its usefulness for cost process planning factors on the OEE. The paper concludes in Section 5
appraisal by appropriately adjusting production time and related time- with managerial insight and theoretical contributions.
dependent cost model elements [17,18]. However, while Fera et al.
[18] acknowledge some differences between AM and the conventional, 2. Theoretical background
repetitive manufacturing context in which OEE was developed, neither
study establishes how OEE should be calculated for AM operations. 2.1. OEE in traditional manufacturing
Attempting to address this gap, Reid [19] develops the Overall Additive
Manufacturing Effectiveness (OAME) indicator. This work offers a useful Originally developed by Seiichi Nakajima [26], OEE is a
insight into the role of in-situ defect mitigation as a production loss, as it well-established equipment performance indicator, based on time and
diminishes time available for effective production. However, the author output wasted via six production losses: breakdown, setup & adjust­
assumes that all defects can be captured and corrected in situ, which is ments, reduced speed, idleness & minor stops, quality defects, and
unrealistic. In consequence, the OAME metric is incomplete as other reduced yield. These are condensed into three metrics, which focus on
quality issues are neglected, such as non-correctable production defects, equipment utilization (availability), the operational rate (performance),
one of the four defect categories identified by Baumers and Holweg [20]. and satisfactory output (quality) [27]. Therefore, OEE calculates the
Thus, a comprehensive translation of the OEE metric to account for the fraction of planned production time which results in good quality
characteristics of AM, such as geometric freedom and poor reliability, output. The structure of OEE is summarized in Fig. 1.
remains missing.
Furthermore, the literature is marked by a limited understanding of 2.2. Framework to translate OEE for AM operations
how AM operations affect the production losses and OEE. Gopsill and
Hicks [21] use their own metric to show that the productivity and per­ 2.2.1. Applying OEE within an AM process workflow
formance of each machine depends on how it is operated and managed. The first question in applying OEE to AM relates to the system
However, the OEE-related studies only offer cursory examples of pro­ boundary for OEE measurements. The steps required to produce finished
duction loss drivers, such as lack of machine component synchronisation products in AM, also referred to as the workflow, spans a system of
[17] and time wasted with poor error reporting systems [19]. Given that different pieces of equipment, each dedicated to specific purposes. For
OEE can capture the impact of manufacturing operations management polymeric LS, major tasks in this workflow include: build file prepara­
on the production losses within the equipment [22], this missing link tion at a computer workstation; manufacture at the AM machine; recy­
connecting OEE to AM operations is particularly unfortunate. cling and mixing feedstock material at a material preparation machine;
A particular AM operations challenge is the multi-dimensional pro­ unpacking then cleaning the parts prior to finishing processes; and
cess planning problem of converting incoming orders into build jobs for extracting reusable feedstock at a post-processing machine.
on-time delivery [23]. While this is somewhat analogous to batch sizing OEE is calculated with respect to specific equipment, rather than the
in traditional manufacturing, the fungibility of AM build space (i.e. the entire process workflow [27,28]. Therefore, only the steps that involve
capability of mutual interchange of units of build space) allows capacity the AM machine itself are assessed in this paper, shown in Fig. 2 as grey
to be used more flexibly [24]. This reduces process planning constraints boxes. It is assumed that the build file and material preparation are
and expands the operations ‘design space’ in which to search for the completed correctly using separate equipment.
optimized solutions. There is a spectrum of approaches for AM opera­ Cleaning and regular maintenance are excluded from the OEE
tions to address this problem, referred to as operations approaches in this calculation in traditional manufacturing as these processes fall under
paper, which range from high-variety/single-part builds to planned downtime [16,27]. However, an adjustment in the definition of
low-variety/high-volume builds [20]. The operations approaches in this planned production time is required for AM. In AM, the build chamber in
paper are chosen to compare and contrast opposing ends of the over­ the machine must be cleaned between each build. Time taken for this
arching management trade-off between build volume utilization and step eats into the planned production time depending on the number of
timely delivery, which are critical in higher-volume manufacture [25]. builds. In consequence, this process is directly related to AM process
Therefore, the impact of the AM operations management on the OEE and planning decisions. Hence, machine cleaning between builds is included
production losses can be assessed. With a better understanding of these in the planned production time as a changeover step, not planned
relationships, operations managers may use OEE as a tool to support downtime. Other steps that follow similar reasoning include process
effective AM operations, which is instrumental for its successful appli­ parameter checks on the machine control system, material and con­
cation within any industry context. sumables refill during machine preparation, and automatic pre-print
This paper seeks to address three identified gaps in the literature: the checks during build release.
missing definition of OEE calculation in the AM context, an absence of
empirical OEE estimates for AM, and a lack of guidelines how AM op­ 2.2.2. Adjusting the OEE equations for use in AM processes
erations could be improved through measuring OEE towards increasing Just as the system boundary for production losses requires adjust­
production effectiveness in higher-volume applications. This is achieved ment, the equations used to measure the losses and to calculate OEE
by formulating a framework to systematically adapt OEE for AM and require adaption. Rather than altering the structure of the OEE calcu­
following exploratory simulation-based approach to provide an indica­ lation entirely as in [19], we have opted to retain the OEE structure and
tion of the use of this metric in the operations management context. adapt the equations of the constituent metrics (availability, perfor­
Simulation is chosen to efficiently explore general relationships between mance, quality) to suit the operations characteristics of additive
AM operations and OEE that are proposed for further empirical testing. manufacturing. This approach follows the work of Jauregui Becker et al.
While this paper develops and applies the metric to polymeric Laser [29], who adapt OEE for high-mix low-volume machining.
Sintering (LS), belonging to the powder-bed fusion variant of AM tech­ For conventional manufacturing processes, it is important to note
nology, its generalizability to other forms of AM is also discussed. In that multiple approaches have been proposed for calculating the per­
doing so, this paper aims to stimulate a wider research agenda towards formance and quality metrics. Some authors use purely time-based
leveraging established operations management methods in AM, which is measurement [19,29], whereas others monitor output alongside time
urgently needed. [16,26,28,30]. This paper suggests that a focus on output quantity and
The remainder of this paper is structured as follows. Section 2 pro­ per-unit cycle time is not appropriate for AM. Additionally, per-layer
vides the theoretical background by reviewing the OEE concept and cycle times would be difficult to measure as the production time
proposes a framework for its application in AM. Sections 3 and 4 extend changes depending on the contents of each layer in an AM build. The
this by simulating the effect of three operations approaches and related approach taken in this paper is therefore to replace per-unit time and

2
S. Basak et al. Additive Manufacturing 56 (2022) 102904

Fig. 1. Schematic summary of the ‘six big losses’ (dashed boxes).

Fig. 2. Process workflow for polymer laser sintering.

per-unit output measures with cubic volume-based metrics, according to 2.3. Understanding production losses within the AM context
Eqs. (1)–(3):
OEE distinguishes between value-adding and non-value adding
planned production time − downtime
Availability : A = (1) processes. This paper takes the position that the only value-adding step
planned production time
in the AM process is the material deposition process occurring within the
AM machine, as shown in Fig. 2. While the other steps are necessary for
successfully producing a good-quality part, they do not create or add

total cubic volume of parts


Performance : P = (2)
actual production time × theoretical volumetric processing rate

total cubic volume of parts − cubic volume of defective parts


Quality : Q = (3)
total volume of parts

These equations match the structure of Nakajima’s equations [26], value to the part per se. In the language of OEE, these elements
thus avoiding unnecessary deviation from original OEE metric. More contribute to losses in the value-adding capacity of the machine. Even
importantly, using cubic volume-based output keeps its focus on within the deposition process step, the machine is idle while newly-
value-adding production as a proportion of the planned, physical added layer of material is heated to the required temperature prior to
equipment capability. Using this approach, the calculation is also un­ laser scanning. This illustrates how particular production losses are
affected by part size or geometry and robust to different AM technology inherent in AM, due to the nature of the layer-by-layer process and
variants. required ancillary steps. For the AM technology variant LS, Fig. 3 pro­
vides examples for production losses. Losses that are the consequence of
the layer-by-layer and batch-wise operating principles of AM are marked

3
S. Basak et al. Additive Manufacturing 56 (2022) 102904

structuring the links between OEE and AM operations can act as a source
for important technical and operational cues for operations managers to
reduce production losses.
From the operations perspective, Baumers and Holweg [20] distil the
ways in which AM deviates from conventional manufacturing into five
characteristics of AM operation:

1. Product variety can be produced at close to zero marginal cost;


2. Each build is a vertical batch process with a maximum batch size
determined by the build volume capacity;
3. Volume-driven static economies of scale apply up to full build vol­
ume utilization, but not beyond;
4. Learning curve effects apply to both pre- and post-processing steps;
and
5. The risk of build failure increases with the number of layers
produced.

Table 1 maps the effect of each AM operations characteristic on the


six production losses monitored by OEE. This codifies the relationships
between AM operations and OEE, and can form a foundation for
continuous improvement initiatives within AM for higher-scale
applications.
In this context, it is crucial to note that, due to the far-reaching
Fig. 3. Examples of production losses at the machine in the AM workflow. freedom of geometry inherent to AM [31] the product variety in AM
tends to hinder the scope for standardization. Moreover, the batch na­
with an asterisk. ture of AM operation (in which one build operation is completed after
While the inherent layer- and batch-related losses may never be another) and the inability to continuously expand machine capacity
completely eliminated, the operation of AM equipment also affects beyond the available build volume results in high, unavoidable pro­
production losses, and by extension, the OEE. Therefore, identifying and duction time losses on a per-build basis, for example resulting from
extensive warm-up and cool-down procedures. Any additional builds

Table 1
Mapping of the production losses to the operations characteristics of AM.
Operational Production Loss
Characteristics of
Breakdown Set-up & Adjustment Idling & Minor Stops Reduced Speed Defects & Start-up Yield
AM
Rework

Product variety at Process parameters in Sacrificial parts may


minimal marginal control system may be required to
cost change, depending on the regulate thermal
parts in each build, conditions across
requiring checks each dissimilar product
time(-)[32] geometries (-)[33]
Feedstock material
variety incurs time
penalty for material
changeover in the
machine (-)[32]
Each build is a vertical Build failure can lead Each build incurs a time Each build incurs a time Each build requires Defects may arise in
batch process, to loss of entire batch penalty for pre- and post- penalty for warm up and deposition of blank parts near the
limited by build of parts, as an process steps (-)[18] cool down of build layers prior to bottom of the build
volume capacity interrupted build Periodic time penalty for volume (-)[35] productive phase (-) volume due to less
cannot be restarted feedstock material refresh [34] stable thermal
(-)[34] (-)[18] conditions (-)[36]
Static economies of Higher utilization of space Space-efficient part
scale up to full build in each layer increases the positioning reduces
volume utilization in proportion of productive time taken for laser
each build time to layer-wise idle spot to traverse
time (+)[17,35] between sintered
regions (+)[37]
Learning curve in pre- Operator skill/experience Efficient standard
process and post- affects time to set up operating procedures for
process steps machine (+)[20] manual steps minimize
errors and minor stops (+)
[19]
Risk of build failure Extra builds required to Extra builds required to Likelihood of
increases with accommodate reworked accommodate reworked rework increases
number of layers parts incur a time penalty parts incur a time penalty with number of
for pre- and post-process for warm up and cool layers occupied
steps (-)[34] down of build volume (-) by parts (-)[34]
[34]

(+) or (-) indicate positive or negative effect on OEE, respectively.

4
S. Basak et al. Additive Manufacturing 56 (2022) 102904

required to accommodate rework further exacerbate these losses. experiments than a full factorial approach [46,47].
The impacts of lead time and product variety are examined sepa­
2.4. Synthesis rately via graphical methods and Analysis of Variance (ANOVA).
ANOVA can be applied to estimate the contribution of each main effect
The theoretical background and framework presented up to this and interaction to the variance in the OEE from the experimental data.
point has covered each step for applying OEE to AM. First, the system The contribution is found by dividing the estimate of variance for each
boundary and relevant tasks for OEE within the AM workflow were effect, given by the sum of the squared deviations from the mean,
defined. Second, the equations for OEE were switched to a cubic volume- through the sum of the estimates for all effects [48]. The effects of lead
based measurement to reflect the geometric freedom and fungible build time and part variety on Identical Batch Make-to-Stock AM are not
volume space associated with AM. This approach is more appropriate assessed, as it is assumed that this operations approach is unaffected by
than the time-based, condensed Overall Additive Manufacturing Effec­ day-to-day changes in product demand.
tiveness indicator [19] as it works across AM technology variants and
does not preclude systems lacking in-situ defect correction.
3.2. Model
Following this, the workflow was examined in more detail to identify
inherent losses and link sources of production losses to the operations
The simulation seeks to replicate the nature of an empirical build
characteristics of AM. This elaborates upon previous work investigating
experiment study, as in Baumers and Holweg [34]. A manual imple­
AM operations and system performance, relating build volume utiliza­
mentation of a discrete event simulation (DES) approach is followed, as
tion to cost [20,35] or batch size to throughput [38]. Compared to cost
DES is considered appropriate for the operational level of decision
or throughput, OEE is a more holistic measure as it benchmarks the
making involved in production planning and resource utilization studies
process against its potential capacity and includes effects of process
[49]. Unless otherwise stated, all parts of the model and simulation are
planning decisions beyond the automatic build step.
implemented via spreadsheets.
The concepts and assumptions in this framework will be tested in the
The simulation is based on a single machine fulfilling a production
following sections, which report on simulation experiments to indicate
schedule for sets of orders of two parts: a surgical guide for the knee
the effect of the operations approach and related factors on the OEE. The
(Fig. 4a) and an unmanned aircraft nose, with helical internal structure
nature of relevant trade-offs, for example between lead time and build
(Fig. 4b). The reference polymer LS machine is the EOS Formiga P100,
volume utilization [25], are assessed to guide operations managers to­
which has a build volume measuring 210 mm by 260 mm by 330 mm
wards the impact of process planning decisions on the effective use of
tall. Manufacturing execution is treated as a one-time packing and
the AM system.
scheduling problem, which is a simplified snapshot of the order queue
system seen in make-to-order AM production [50]. Therefore, the steps
3. Simulation method
in this study are order generation, simulation of order fulfilment
including rework, and calculation of OEE terms.
3.1. Overview
The set of orders are generated randomly, and are common across the
order fulfilment using each operations approach. An example is given in
This section presents the method of a limited, exploratory simulation
Table 2, showing orders to be delivered across three days, corresponding
of the production workflow on a polymeric LS system. The method of
to a lead time of 72 h.
exploratory simulation is applied in manufacturing capability studies to
The order fulfilment process is then simulated (Fig. 5). The order
reveal general relationships about the system under investigation in an
parts are packed into the build volume, starting with any leftover parts
efficient manner, providing insights for focused empirical testing in the
from the previous day. Packing is completed via a research tool,
future [39,40]. In this study, the purpose is to explore the impact of
3DPackRAT (as described in [51]), which automatically implements a
operations approaches and other process planning factors, such as lead
front-bottom-left heuristic. Once a build is packed, the build time is
time and product variety, on the effectiveness and efficiency of AM
estimated using a regression model (Eq. (4)) from the data obtained in
production, as per the OEE. The simulation is therefore structured in two
3DPackRAT. This regression model is formulated from 14 samples of
stages, as follows.
prior empirical build experiment data for the same LS machine, as re­
In the first stage of the simulation study, the OEE and constituent
ported in [20].
metrics for three operations approaches are compared. The literature
( )
identifies three major clusters of AM usage, which form the operations build time = 0.54 + 0.0050lbuild + 4.9 × 10− 6 Vbuild (4)
approaches. The first is Identical Batch Make-to-Stock AM, in which
identical parts are made in fixed, standardized batches that fill the where:
machine space as far as possible to replenish inventory stock [41,42].
The second is Capacity Maximizing Make-to-Order AM, in which lbuild – number of layers in the build
mixed-part batches fill the available machine space to meet a target Vbuild – volume of parts in the build (mm3)
build volume utilization, prioritizing high productivity in fulfilling
incoming orders [9,43]. The third is Lead Time Minimizing Make-to-Order For each build, the cumulative build height and quantity of parts are
AM, which focusses on mixed-part batches where the quantity of parts is checked against the Mean Time Between Failure (MTBF) for outright
capped to limit the process make-span, prioritizing faster delivery for the build failure or defective parts (Table 3). If the MTBF is exceeded, the
incoming orders [44]. part or entire build becomes rework and the cumulative count resets. In
In the second stage, the simulation assesses the effect of lead time Identical Batch Make-to-Stock AM, defects are absorbed by inventory
changes and product variety on OEE for the two make-to-order opera­ stock. Additionally, in the case of outright build failure, the process time
tions approaches, Capacity Maximizing and Lead Time Minimizing for machine warm up, build up to the failed layer, and cool down are
Make-to-Order AM. These factors are chosen as they are considered converted to breakdown time. In the above sequence of steps, the op­
among the distinguishing advantages of AM [45]. A Design of Experi­ erations approach dictates when a new build is triggered (see grey step
ments (DoE) approach is followed for these simulations, covering the in Fig. 5a and b), as described:
three factors of operations approach, lead time, and product variety. The
Central Composite Design Face-Centred design is chosen to fully 1. Identical Batch Make-to-Stock AM – identical parts fill standardized
examine the influence of the factors (main effects, linear and quadratic) builds, each containing the maximum quantity that fit in the build
and their two-way interactions on the OEE with fewer simulation volume;

5
S. Basak et al. Additive Manufacturing 56 (2022) 102904

Fig. 4. Test components used in the simulation.

3.3. Assumptions and Simplifications


Table 2
Example of a set of orders.
Assumptions in the simulation arise during the process of converting
Part Quantity the uncertainties from real-world manufacturing to the model domain.
Day 1 Day 2 Day 3 TOTAL Based on the approach taken by Baumers and Holweg [34] towards AM
A 2 3 4 9 failure modes, this study makes the following two assumptions related to
B 7 5 4 16 breakdown and defect production losses:

• Each build either completes successfully or experiences outright


2. Capacity Maximizing Make-to-Order AM – mixed part builds are build failure, leading to the loss of all parts.
configured, where the quantity matches each set of orders, and each • Failed builds and defective parts are repeated to ensure that the order
build is allowed to occupy the full build height of 330 mm; is completed.
3. Lead Time Minimizing Make-to-Order AM – as for Capacity Maxi­
mizing Make-to-Order AM, but the build height is capped at 100 mm. Additionally, defective parts are always chosen at random from the
According to the build time estimator, Eq. (4), this build height is the original size versions of parts A and B. This ensures that the rework
limit for completing production of up to 500 cm3 of parts within 24 h volume is consistent across the simulation experiments.
(including machine warm up and cool down). Two further assumptions arise relating to the order generation and
fulfilment. First, in each set of orders, the part volume ordered daily is
The OEE is then evaluated for each set of orders. The actual pro­ capped at 500 cm3. This constraint ensures that Lead Time Minimizing
duction time and production losses are estimated, as per Table 3, for Make-to-Order AM is able to deliver on time within 24 h. Second, each
each build in the sequence; and the necessary terms in Eqs. (1)–(3) are build is capped at 10% build volume utilization to match realistic levels
calculated. As for the build time regression model, all fixed time values of machine usage [20].
are taken from prior empirical build experiment data, as in [20]. The A number of simplifications are applied to constrain the detail and
theoretical volumetric processing rate is an estimate of the maximum scope of the simulation performed for this research. To reduce the
hourly output volume for the value-adding deposition step. Using the required detail, the part variety is modelled simply as a difference in the
theoretical maximum build rate of 20 mm/hour [52] and assuming part size, corresponding to 50% or 150% of the original part volume
maximum 10% build volume utilization up to the full build height, the (Fig. 4c). Therefore, the part quantity can be controlled to easily
value given is 109.2 cm3/hour. The simulation and calculation process maintain an unchanged total order volume across the simulation ex­
runs for one day’s operations and is repeated for five sets of orders, or periments. To limit the scope of the simulation, two of the four identified
repetitions over one working week. This aligns with modes of AM failure [20] are omitted. The first, correctable post-build
industry-recommended daily OEE measurement frequency, alongside part rejection, is neglected as any rework is completed away from the
weekly summaries [15].

Fig. 5. Flowchart of the order fulfilment process for each set of orders for a) Identical Batch Make-to-Stock AM, and b) Capacity Maximizing and Lead Time
Minimizing Make-to-Order AM.

6
S. Basak et al. Additive Manufacturing 56 (2022) 102904

Table 3
Summary of parameters contributing to production losses in AM.
Production Loss Value Source

Breakdown Mean time between failure (build failure): 6244 layers [20]
Set-up & Adjustment Fixed pre-process time: 0.25 h Prior empirical data, used in[20]
Fixed post-process time: 0.23 h Prior empirical data, used in[20]
Idling & Minor Stops Fixed warm up time: 3.51 h Prior empirical data, used in[20]
Fixed cool down time: 12 h Prior empirical data, used in[20]
Theoretical volumetric processing rate: 109.2 cm3/hour Calculation based on data from[52]
Reduced Speed Theoretical volumetric processing rate: 109.2 cm3/hour Calculation based on data from[52]
Defects & Rework Mean time between failure (non-correctable part rejection): 40 units [20]
Start-up Yield Ignored in this study Simplification: thermal conditions in build volume are not modelled

AM machine and this does not affect the equipment OEE; and the sec­ availability by only 1%. On the other hand, availability is up to 35%
ond, mechanical or material failure, is excluded as it is assumed that lower for the simulation runs where breakdown occurs in Identical
material preparation is completed successfully (see Section 2.2.1). Batch Make-to-Stock AM and Capacity Maximizing Make-to-Order AM,
compared to 17% in Lead Time Minimizing Make-to-Order AM. This
3.4. Experiment design difference relates to the build height. When build failure occurs, the time
for machine warm up, build up to the failed layer, and cool down are
For the first stage of the simulation, the OEE is obtained for each wasted. In Identical Batch Make-to-Stock AM and Capacity Maximizing
operations approach based on a default lead time of 72 h for each Make-to-Order AM, as the builds are taller i.e. exhibit a greater height
component. This lead time level is seen as the upper limit of responsive and more layers are deposited in the failed builds, more time is lost to
make-to-order AM production [2,44]. breakdown. Consequently, the overall mean and standard deviation
For the second stage of the simulation, the DoE design, Central (error bars in Fig. 6) of the availability are both poorer for these two
Composite Design Face-Centred, is implemented with three factors at operations approaches, indicating lower confidence in consistent
three levels, as per Table 4. The factor Allowable Build Height (ABH), is a equipment uptime. As the setup time per build is much shorter than the
proxy for the operations approach, referencing the difference in the breakdown losses, many emptier builds (to reduce breakdown losses)
height of the allowed build slice between the operations approaches. are favoured over fewer, fuller builds (to reduce setup losses); and
The choice of a continuous variable proxy allows for three levels in this therefore, Lead Time Minimizing Make-to-Order AM is beneficial here.
factor within the DoE, which is a requirement to examine the interaction The performance metric illustrates the penalty of an increased
terms, without introducing another redundant operations approach to number of builds, resulting from idle time during machine warm up and
the analysis. The Lead Time (LT) factor controls the ‘length’ of the order cool down. This loss dominates at 40% of actual production time for
queue snapshot by altering the number of orders available to fulfil. Identical Batch Make-to-Stock AM, 54% for Capacity Maximizing Make-
Finally, smaller and larger varieties of the sample parts are introduced to-Order AM, and 69% for Lead Time Minimizing Make-to-Order AM.
(Fig. 4c); the factor Part Size Variety (PSV), represents the range in size Moreover, in the extreme case of a single-part build, for example, an
across the parts in the orders, relative to the original part volume. The ‘overflow’ build that only contains the individual replacement for a
order quantities are adjusted such that the total geometric volume to be defective part, the idle time rises to four-fifths of the actual production
produced is unchanged in order to isolate the effect of variety in size. time. The impact of the excess idle time on the performance metric is
particularly significant for the Capacity Maximizing Make-to-Order AM
4. Simulation results approach, whereby the mean is 22% higher when single-part builds are
excluded in the analysis.
4.1. Results for operations approach Focusing on the deposition step, the volume deposited per unit build
time indicates that reduced speed losses are lowest in Identical Batch
Fig. 6 reports the mean values for OEE and constituent metrics over Make-to-Stock AM. The build volume utilization here is consistently
five repetitions in the simulation. Adopting the Identical Batch Make-to- 10%. This indicates the importance of consistent and efficient part
Stock AM operations approach results in the highest OEE (35%), fol­ packing for process speed. Other potential sources of performance losses
lowed by Capacity Maximizing Make-to-Order AM (24%), and Lead include accumulating layer-wise idle time in the taller Identical Batch
Time Minimizing Make-to-Order AM (16%). Identical Batch Make-to-
Stock AM incurs the smallest production losses; this is because filling
each build maximizes the batch capacity and takes advantage of static
economies of scale in each build.
As expected, machine availability benefits from the adoption of the
Lead Time Minimizing Make-to-Order AM approach. While the setup
losses are higher in line with the number of builds, this impacts the

Table 4
Factors for simulation via design of experiments.
Level Factor

Allowable Build Lead Time Part Size Variety (%)


Height (mm) (hours)

1 100 48 0
2 215 72 50 (50%, 100% volume
parts)
3 330 96 100 (50%, 100%, 150%
volume parts)
Fig. 6. OEE and constituent metrics for the investigated operations approaches.

7
S. Basak et al. Additive Manufacturing 56 (2022) 102904

OEE for Lead Time Minimizing Make-to-Order AM and Capacity Maxi­


mizing Make-to-Order AM. From 16 experiments, Fig. 7 shows the main
effects (linear and quadratic) and two-way interaction terms that are
significant at the 10% confidence interval, while observing model hi­
erarchy. The main effects (Fig. 7a) are the influence of a single factor,
operations approach (Allowable Build Height) or lead time or part size
variety, on the OEE. The interactions (Fig. 7b) illustrate how the value of
one factor influences the effect of a second factor on the OEE; the only
statistically significant interaction is between the operations approach
(Allowable Build Height) and the lead time.
First, the lead time main effect in Fig. 7a (middle panel) shows that
extending the order lead time from 48 h to 96 h improves the OEE by
2.5% on average. With a longer lead time, a greater quantity of parts is
available to pack, and so the volume deposited per build rises. This
improves the ratio of build time to idle time and reduces performance
losses.
However, the effect of lead time is not uniform across the operations
strategies, indicated by the non-parallel curves in Fig. 7b, which arise
from the interaction with the operations approach. The curves converge
at 100 mm allowable build height, equivalent to Lead Time Minimizing
Make-to-Order AM; this is because the volume deposited and number of
builds increase at a similar rate, leading to negligible improvement in
the performance metric or OEE. Whereas, at 330 mm, equivalent to
Capacity Maximizing Make-to-Order AM, the volume deposited per
build is 54% higher, and the OEE jumps by 5%. Therefore, it is important
to control both the lead time and operations approach for the most
efficient use of the available machine space.
The main effects curve for Allowable Build Height (Fig. 7a, left
panel) also show the dominant effect of the operations approach, as the
change in the OEE of 8% across the three levels is the largest of the three
Fig. 7. Sensitivity of OEE to a) allowable build height, order lead time and part
factors investigated. Moreover, the maximum point in the same curve at
size variety, and b) statistically significant two-way interactions. approximately ABH = 295 signifies the trade-off between increasing
availability losses and decreasing idle losses as the build height in­
creases, as discussed in the previous section.
Make-to-Stock AM and Capacity Maximizing Make-to-Order AM builds,
The main effect of the part size variety (Fig. 7a, right panel) shows
and sparsely filled layers containing the tallest portions of parts in Lead
that switching from no size variety to a mix of 50% and 100% volume
Time Minimizing Make-to-Order AM. Nevertheless, minor changes in
parts, equivalent to PSV = 50, leads to a drop of 3% in the OEE. Thus,
the build time are dwarfed by the idle losses from the warm up and cool
the performance and OEE are sensitive to the variety in part size due to
down steps; and therefore, with respect to performance, the Identical
its effect on the space-efficiency of part packing. It is observed that the
Batch Make-to-Stock AM, with the fewest builds, results in the higher
volume deposited per layer falls by 10%, and so reduced speed losses
OEE observed.
accumulate here. The OEE value subsequently recovers when builds
Quality losses are consistent across the operations approaches, and
contain a mix of 50%, 100% and 150% volume parts, equivalent to
the defect rate of 2.5% is much smaller than the availability and per­
PSV = 100. In this case, the smaller and larger parts fit together more
formance losses. Nevertheless, there is a knock-on effect for availability
favourably within the build volume and they are packed as tightly, on
and performance losses as additional layers and builds are added,
average, as the homogeneously sized parts.
increasing the risk of breakdown and sparsely-filled machine space as
ANOVA is used to compare the relative contributions of the opera­
mentioned above. Therefore, rework has a wider impact than simply
tions approach (proxied by Allowable Build Height), lead time and part
repeating the manufacture of a defective part.
size variety, including quadratic and interaction terms, to the change in
OEE. The contribution is found by dividing the sum of squares (Adj SS),
4.2. Results for lead time and part variety sensitivity an estimate of the variance, by the total sum of squares. The results are
shown for the statistically significant terms in Table 5. The main effect of
To elaborate further, two key parameters in AM process planning – the lead time is lowest, at 2% of the change in OEE, and the part size
lead time and part variety – are changed to illustrate their impact on variety follows at 5%, including the quadratic term. However, the

Table 5
ANOVA for model assessing the sensitivity of OEE.
Model Term DF Adj SS Adj MS F-Value P-Value % Contribution

Blocks 4 0.081 0.020 10.93 0.000 25%


Allowable Build Height (ABH) 1 0.084 0.084 45.62 0.000 26%
Lead Time (LT) 1 0.008 0.008 4.18 0.045 2%
Part Size Variety (PSV) 1 0.000 0.000 0 0.944 0%
ABH × ABH 1 0.013 0.013 7.28 0.009 4%
PSV × PSV 1 0.017 0.017 9.07 0.004 5%
Interaction: ABH, LT 1 0.005 0.005 2.83 0.097 2%
Error 69 0.128 0.002 39%
Total 79 0.326

8
S. Basak et al. Additive Manufacturing 56 (2022) 102904

operations approach remains the dominant effect on the OEE, at 26%, such as the saw-tooth nature of production cost whereby fully occupying
corroborating the findings from Fig. 7a. Therefore, efforts to improve the available slices of build volume is most cost-efficient and there is a fixed-
OEE would be most effective if focused on the operations approach, cost increment for new builds [35]. However, it is important to note that
while the lead time and part size variety offer a moderate level of the size and geometry of the parts in the build affects how efficiently
control. they can be packed [59]. In this study, the chosen test parts coinciden­
tally produce well-packed builds for Identical Batch Make-to-Stock AM,
5. Discussion and conclusions with build volume utilization close to 10%. Whereas, in practice, the
Identical Batch Make-to-Stock AM approach may have to be altered to
5.1. Framework for OEE in AM and production losses allow a standard batch of mixed parts that maximizes utilization of the
machine capacity for a superior performance metric. Of the two flexible
First and foremost, our study confirms that the complete OEE metric operations approaches that allow mixed-part production, Capacity
– with some conceptual adjustments – can be applied to AM processes. Maximizing Make-to-Order AM achieves better build volume utilization
The main message of this paper is that the fundamental logic of targeting and reduces the number of builds relative to the output produced.
the six production losses is just as important in AM as it is in tool-based Therefore, associated setup and idle losses are lower here.
manufacturing. The simulation results corroborate the mapping between the AM
More specifically, the relationships shown in Table 1 demonstrate process characteristics and production losses proposed in Section 2.3,
that all six production losses apply to AM. This contrasts with Fera et al. particularly relating to the batching, static economies of scale, and build
[18], who state that there are no reduced speed losses as the actual failure characteristics. In Identical Batch Make-to-Stock AM and Ca­
process time matches the predicted time, and no start up yield losses as pacity Maximizing Make-to-Order AM, a trade-off is observed between
the AM build process is uniform. These assertions conflict with the increasing breakdown losses and decreasing setup and idle losses in the
available literature, cited in Table 1 and as follows. Across 10 empirical taller, fuller builds. This reflects the cost relationship between build
build experiments on the same machine with identical setup and part failure and economies of scale, whereby rising risk-related costs over­
packing, reported in [20], a range of 55 min (37%) is observed in the power indirect cost efficiencies as capacity utilization increases [34]. On
build time. This indicates that the cycle times for steps in the AM build the other hand, the additional single-part builds for rework experience
process are not constant relative to predictions. Regarding the yield no scale economies and occupy the peaks of the aforementioned
losses, layer-wise thermal control challenges are identified for laser saw-tooth cost curve [35]. The space-efficient packing for zero part size
sintering based on part geometry and deposition over powder or sintered variety and the mix of 50%, 100% and 150% volume parts, as indicated
regions [53]. Therefore, the build time and part quality is not consistent by the volume deposited per layer, reduces performance losses. This is
within individual builds or across builds, and so AM can experience both related to the laser scan path-based build time relationship formulated
reduced speed and yield losses, contradicting Fera et al. [18]. by Pham and Wang [37], whereby the distribution of the sintered areas
In the translation framework (Section 2.2), Eqs. (1)–(3) are presented across the build volume affects the total time.
as appropriate conversions of the original OEE component metrics. The
approach follows Jauregui Becker et al. [29] in retaining the original 5.3. Limitations
OEE structure whilst making adjustments for the manufacturing system
characteristics. The equations for performance and quality are similar to While our study contributes to the growing evidence base of how
rate factor and yield, respectively, which are components of the quantity, quality and cost relate in the AM process under production
’equipment effectiveness’ metric [22]. However, the cubic volume conditions, there are limitations to the findings. In constraining the
frame of reference in this study is better suited to the geometric freedom scope of this exploratory study, only single-machine AM operations are
and fungible build space [24] in AM. investigated, whereas flexible AM production centres require multiple
The OEE framework also improves the breadth of quality issues AM machines operating in parallel. Moreover, the use of two test part
covered, as compared to the OAME metric developed by Reid [19]. In geometries limits the comparison of build volume packing and, by
the OAME calculation, the time lost to in-situ defect mitigation falls extension, the performance metric across the operations approaches.
under performance losses, which is a logical assumption as the ideal Restrictions in the scope of the OEE metric itself must also be
running speed is diminished [16,29], and no other quality issues are acknowledged. Despite its strength as a simple yet comprehensive
monitored. However, while in-situ systems for closed-loop LS control metric, OEE does not explicitly monitor how well equipment integrates
have been developed for thermally-induced part defects [54,55], it is not with the wider manufacturing flow [13]. In contrast, it is argued that the
yet possible to correct all quality issues, such as part shifting or surface OEE is affected by interruptions to the inflow of upstream
defects [56–58]. Therefore, parts may be produced with unacceptable work-in-progress [22] and manufacturing fluctuations such as the
quality, and OAME would not reflect this. In contrast, the performance product mix [60], as the results of this study also confirm. Nevertheless,
metric in this study can capture reduced speed losses relative to the expanded manufacturing performance metrics, such as Overall
theoretical volumetric processing rate for defect mitigation; and the Throughput Effectiveness and Overall Asset Effectiveness, can monitor
quality metric captures any post-production rejects. Taken together, production losses across a production facility better because they cap­
these aspects suggest that the OEE formulation proposed in this study is ture interdependencies between equipment and a wider range of in­
more generalizable than the OAME metric previously suggested by Reid efficiencies, respectively [15,61]. Other extensions to the OEE metric
[19]. increase its coverage of value-adding and non-value-adding activities,
for example, timely customer delivery (Global Process Effectiveness),
5.2. Simulation Study resource consumption (Overall Resource Effectiveness), and planned
maintenance (Total Effective Equipment Performance) [15,62,63].
Our simulation results show that the approach to operating the AM is These adaptations improve the link between OEE and profitability of the
a strong determinant of its performance in terms of OEE achieved. process, without compromising the conciseness of the metric.
Specifically, the results suggest that Identical Batch Make-to-Stock AM The simulation method chosen for this study also has limitations. It is
outperforms the alternative operations approaches in terms of OEE not possible to capture the empirical, random variation within the AM
achieved. This is because setup and idle losses associated with switching build process using modelled elements such as the regression-based es­
from one build to the next are minimized, and the invariant, space- timate of build time, as discussed in the previous section. From an in­
efficient packing of the chosen test parts maintains lower reduced dustry application perspective, the operator time and effort for
speed losses. Quantity-cost relationships for AM exhibit a similar logic, recording the data required for OEE measurement has also been

9
S. Basak et al. Additive Manufacturing 56 (2022) 102904

neglected. In reality, automated solutions for gathering and categorising While these measures could adversely affect the OEE of each indi­
data may be developed, albeit with upfront implementation time and vidual machine, the lead time and resource-related operational losses
cost [19]. Nevertheless, applying a simulation-based methodology for across the entire AM workflow would reduce. For example, kitted parts
hypothetical problem-solving using artificial data can generate useful may sub-optimally fill the build volume, leading to performance losses,
generic insights [49], in this case, for AM operations management. whereas reducing the post-build workflow steps improves delivery lead
time [51]. Alongside the expanded metrics discussed in the previous
5.4. Managerial insights section, monitoring production cost and related metrics such as Opera­
tional Excellence Profitability can provide further financially oriented
In order to maximize process efficiency, AM users should in principle insights into AM machine operations [20,60], albeit with more onerous
strive for the operations approach with the highest OEE, which corre­ data acquisition and calculations than OEE. Overall, manufacturing
sponds to low levels of the six production losses. Moreover, AM pro­ performance metrics provide an important linking point between
duction cost decreases alongside production losses, depending on manufacturing equipment and Enterprise Resource Planning (ERP)
utilisation (availability) and build rate (performance) [9,64]. Our results systems and dashboard implementations aiming to provide visibility of
have strong managerial relevance as they prescriptively guide AM op­ the performance of physical capital. This continuous production
erations managers on how to improve OEE when the operations floor-based monitoring, in itself, can drive improvement initiatives [13].
approach requires variety and flexibility. Further research is also required to generalise the OEE framework for
First, we show that to maximize OEE in AM, the foremost priority other AM techniques, such as metal powder bed fusion processes and
should be to minimize the number of builds. This will increase the ratio stereolithography, among other candidates for higher volume AM pro­
of productive deposition time to inherent and expensive losses occurring duction. One area of interest will be support structures, which are
per build, chiefly machine warm up and cool down. Yet, the adverse necessary in both aforementioned processes. Support structures are
consequence is to increase time lost to breakdown in the case of build likely to lead to additional production losses including reduced speed,
failure, which may lead to unacceptable further consequences, such as analogous to sacrificial parts in polymer powder bed fusion, and start-up
late delivery and wasted overhead costs during downtime. yield, as unrecoverable material is deposited here to stabilise the parts
Second, the allocation and packing of parts should seek to achieve being built.
high build volume utilization and consistent distribution of size and In summary, the proposed framework links the six production losses
shape in the parts, appropriate to the AM machine. For the LS machine to characteristics of AM and shows how the AM operations approach
and test parts in this study, homogeneous part size and shape are most impacts on these. This extends previous efforts to adapt OEE for high-
favourable, but similar space-efficient packing is achieved when mixed- mix-low-volume conventional manufacturing [29], by applying a
size parts fit well together in a given build volume. This minimizes un­ similar approach to develop suitable OEE equations for AM, which in­
certainty in reduced speed losses and improves the confidence in the volves much greater variety in the process operations and output parts.
performance metric recorded. The OEE metric proposed here offers a broader scope than the OAME
Third, efforts to extend the order lead time will reduce production metric [19], as the structure for calculating production losses is more
losses, when applied in conjunction with the previous two points. With generalizable for AM technologies and, unlike OAME, incorporates de­
the Capacity Maximizing Make-to-Order AM approach, increasing the fects that cannot be corrected in-situ. We build on the work of Fera et al.
lead time improves the likelihood of filling each build volume and thus [18] via our simulation study and discussion, which provide an estimate
achieving a better ratio of productive build time to idle time lost. It is of the OEE for the chosen LS machine and illustrate that all six pro­
important to acknowledge, however, that this is likely to be outside AM duction losses are present in AM. Most importantly, however, we pro­
managers’ control, as the short lead time is a value-adding feature of vide AM operators with a much-needed structured approach to assess
AM. the effectiveness with which their AM equipment is used.
Overall, this paper highlights that there is considerable room for
improvement in the OEE of laser sintering, even in Identical Batch Make- Funding disclosure
to-Stock AM where the OEE is estimated at 35%. While no comparable
empirical estimates for AM are available, the results in this study are at This work was supported by Engineering and Physical Sciences
the lower end of the range found in conventional manufacturing pro­ Research Council [grant number: EP/L01534X/1].
cesses, which is 30–80% [13]. The constituent metrics strongly indicate
that attention should be placed on performance, and so future empirical
CRediT authorship contribution statement
work would focus on the impact of technological advances in AM pro­
cesses, such as the introduction of High Speed Sintering processes [65,
Shreeja Basak: Conceptualization, Methodology, Investigation,
66]. Additionally, the ability to cool down one build while simulta­
Formal analysis, Visualization, Writing − original draft. Martin
neously starting the next would significantly improve idle losses [67,
Baumers: Conceptualization, Methodology, Writing − review & editing,
68]; as would parallelization in the layer-by-layer deposition process,
Supervision. Matthias Holweg: Conceptualization, Methodology,
such as through the adoption of multiple energy sources [10]. The extent
Writing − review & editing. Richard Hague: Supervision, Funding
to which these technological changes improve the OEE is yet to be
acquisition. Chris Tuck: Supervision, Funding acquisition.
investigated.
Moreover, each machine exists within a wider ecosystem of pro­
duction, including part finishing and further manufacturing steps away
Declaration of Competing Interest
from the immediate AM workflow (Fig. 2); and successful balance and
management integration within this expanded manufacturing system is
The authors declare that they have no known competing financial
seen as key for AM implementation [23]. Beyond the operations ap­
interests or personal relationships that could have appeared to influence
proaches explored in this work, other approaches could be introduced to
the work reported in this paper.
improve the system balance, for example, dynamic scheduling and
staggering of builds to avoid queuing of parts for post-processing, or to
avoid builds finishing outside operator shift hours. Additionally, parts References
can be grouped by orders they belong to, an approach also known as
[1] I. Gibson, D. Rosen, B. Stucker. Additive Manufacturing Technologies: 3D Printing,
‘kitting’, to ease post-build assembly time, labour and inventory re­ Rapid Prototyping and Direct Digital Manufacturing, Second edition.,, Springer,
quirements [51]. New York Heidelberg Dodrecht London, 2015.

10
S. Basak et al. Additive Manufacturing 56 (2022) 102904

[2] D. Deradjat, T. Minshall, Implementation of rapid manufacturing for mass [30] S.H. Huang, J.P. Dismukes, J. Shi, Q. Su, M.A. Razzak, R. Bodhale, D.E. Robinson,
customisation, J. Manuf. Technol. Manag. 28 (2017) 95–121, https://doi.org/ Manufacturing productivity improvement using effectiveness metrics and
10.1108/JMTM-01-2016-0007. simulation analysis, Int. J. Prod. Res. 41 (2003) 513–527, https://doi.org/
[3] Siemens, 3D-printing of custom-made shoe sole, Siemenscom Glob. Website, 2020. 10.1080/0020754021000042391.
〈https://new.siemens.com/global/en/company/stories/industry/am-additiveman [31] R. Hague, S. Mansour, N. Saleh, Design opportunities with rapid manufacturing,
ufacturing-3dprinting-textile-eos-dyemansion.html〉 (accessed 18 May 2021). Assem. Autom 23 (2003) 346–356, https://doi.org/10.1108/
[4] GE Additive, GA-ASI completes first test flight with metal 3D-printed part | GE 01445150310698643.
Additive, 2020. 〈https://www.ge.com/additive/press-releases/ga-asi-completes-fi [32] H. Proff, A. Staffen, Challenges of Additive Manufacturing: Why companies don’t
rst-test-flight-metal-3d-printed-part〉 (accessed 18 May 2021). use Additive Manufacturing in serial production, Deloitte, 2019. 〈https://www2.de
[5] S. Zanoni, M. Ashourpour, A. Bacchetti, M. Zanardini, M. Perona, Supply chain loitte.com/content/dam/Deloitte/de/Documents/operations/Deloitte_Challenges_
implications of additive manufacturing: a holistic synopsis through a collection of of_Additive_Manufacturing.pdf〉 (accessed June 23, 2021).
case studies, Int. J. Adv. Manuf. Technol. 102 (2019) 3325–3340, https://doi.org/ [33] M. Pavan, M. Faes, D. Strobbe, B. Van Hooreweder, T. Craeghs, D. Moens,
10.1007/s00170-019-03430-w. W. Dewulf, On the influence of inter-layer time and energy density on selected
[6] S.H. Khajavi, J. Holmström, J. Partanen, Additive manufacturing in the spare parts critical-to-quality properties of PA12 parts produced via laser sintering, Polym.
supply chain: hub configuration and technology maturity, Rapid Prototyp. J 24 Test. 61 (2017) 386–395, https://doi.org/10.1016/j.polymertesting.2017.05.027.
(2018) 1178–1192, https://doi.org/10.1108/RPJ-03-2017-0052. [34] M. Baumers, M. Holweg, Cost Impact of the Risk of Build Failure in Laser Sintering,
[7] H.-J. Steenhuis, L. Pretorius, The additive manufacturing innovation: a range of in: Proc. Solid Free. Fabr. Symp. 2016 Univ. Tex. Austin, Austin, Texas, 2016: pp.
implications, J. Manuf. Technol. Manag. 28 (2017) 122–143, https://doi.org/ 1910–1921.
10.1108/JMTM-06-2016-0081. [35] M. Ruffo, C. Tuck, R. Hague, Cost estimation for rapid manufacturing - laser
[8] R. D′ Aveni, The 3-D Printing Revolution, Harv. Bus. Rev., 2015. 〈https://hbr.org/ sintering production for low to medium volumes, Proc. Inst. Mech. Eng. Part B J.
2015/05/the-3-d-printing-revolution〉 (accessed 26 February 2018). Eng. Manuf 220 (2006) 1417–1427, https://doi.org/10.1243/09544054JEM517.
[9] M. Baumers, P. Dickens, C. Tuck, R. Hague, The cost of additive manufacturing: [36] A. Wegner, G. Witt, Understanding the decisive thermal processes in laser sintering
machine productivity, economies of scale and technology-push, Technol. Forecast. of polyamide 12, in: Proc. PPS-30 30th Int. Conf. Polym. Process. Soc. – Conf. Pap.,
Soc. Change. 102 (2016) 193–201, https://doi.org/10.1016/j. Cleveland, Ohio, USA, 2015: p. 160004. 〈https://doi.org/10.1063/1.4918511〉.
techfore.2015.02.015. [37] D.T. Pham, X. Wang, Prediction and reduction of build times for the selective laser
[10] EOS GmbH, EOS LaserProFusion for Tool-Free Injection Molding | EOS, 2021. 〈http sintering process, Proc. Inst. Mech. Eng. Part B J. Eng. Manuf 214 (2000) 425–430,
s://www.eos.info/en/innovations/3d-printing-of-the-future/laserpro-fusion〉 https://doi.org/10.1243/0954405001517739.
(accessed 26 August 2021). [38] T. Stittgen, J.H. Schleifenbaum, Towards operating curves of additive
[11] R.A. D′ Aveni, A. Venkatesh, How to Make 3D Printing Better, Harv. Bus. Rev., manufacturing equipment: production logistics and its contribution to increased
2020. 〈https://hbr.org/2020/09/how-to-make-3d-printing-better〉 (accessed 27 productivity and reduced throughput time, Procedia CIRP 93 (2020) 14–19,
January 2021). https://doi.org/10.1016/j.procir.2020.06.006.
[12] K.K.B. Hon, Performance and evaluation of manufacturing systems, CIRP Ann 54 [39] A. Größler, An exploratory system dynamics model of strategic capabilities in
(2005) 139–154, https://doi.org/10.1016/S0007-8506(07)60023-7. manufacturing, J. Manuf. Technol. Manag 21 (2010) 651–669, https://doi.org/
[13] P. Jonsson, M. Lesshammar, Evaluation and improvement of manufacturing 10.1108/17410381011063978.
performance measurement systems - the role of OEE, Int. J. Oper. Prod. Manag. 19 [40] R. Ramasesh, S. Kulkarni, M. Jayakumar, Agility in manufacturing systems: an
(1999) 55–78, https://doi.org/10.1108/01443579910244223. exploratory modeling framework and simulation, Integr. Manuf. Syst 12 (2001)
[14] N. Kang, C. Zhao, J. Li, J.A. Horst, A Hierarchical structure of key performance 534–548, https://doi.org/10.1108/EUM0000000006236.
indicators for operation management and continuous improvement in production [41] G. Avventuroso, R. Foresti, M. Silvestri, E.M. Frazzon, Production paradigms for
systems, Int. J. Prod. Res. 54 (2016) 6333–6350, https://doi.org/10.1080/ additive manufacturing systems: A simulation-based analysis, in: 2017 Int. Conf.
00207543.2015.1136082. Eng. Technol. Innov. ICEITMC, IEEE, Funchal, 2017: pp. 973–981. 〈https://doi.
[15] P. Muchiri, L. Pintelon, Performance measurement using overall equipment org/10.1109/ICE.2017.8279987〉.
effectiveness (OEE): literature review and practical application discussion, Int. J. [42] N. Hopkinson, P. Dickens, Analysis of rapid manufacturing—using layer
Prod. Res. 46 (2008) 3517–3535, https://doi.org/10.1080/00207540601142645. manufacturing processes for production, Proc. Inst. Mech. Eng. Part C J. Mech.
[16] B. Dal, P. Tugwell, R. Greatbanks, Overall equipment effectiveness as a measure of Eng. Sci 217 (2003) 31–39, https://doi.org/10.1243/095440603762554596.
operational improvement – A practical analysis, Int. J. Oper. Prod. Manag. 20 [43] M. Ruffo, R. Hague, Cost estimation for rapid manufacturing - simultaneous
(2000) 1488–1502, https://doi.org/10.1108/01443570010355750. production of mixed components using laser sintering, Proc. Inst. Mech. Eng. Part B
[17] S. Dirks, J. Schleifenbaum, Adaption of cost calculation methods for modular laser- J. Eng. Manuf. 221 (2007) 1585–1591, https://doi.org/10.1243/
powder bed fusion (L-PBF) machine concepts, in: Proc. Met. Addit. Manuf. Conf. 09544054JEM894.
2019, Orebro, Sweden, 2020: pp. 79–87. [44] A. Chergui, K. Hadj-Hamou, F. Vignat, Production scheduling and nesting in
[18] M. Fera, F. Fruggiero, G. Costabile, A. Lambiase, D.T. Pham, A new mixed additive manufacturing, Comput. Ind. Eng. 126 (2018) 292–301, https://doi.org/
production cost allocation model for additive manufacturing (MiProCAMAM), Int. 10.1016/j.cie.2018.09.048.
J. Adv. Manuf. Technol. 92 (2017) 4275–4291, https://doi.org/10.1007/s00170- [45] M. Attaran, The rise of 3-D printing: The advantages of additive manufacturing
017-0492-x. over traditional manufacturing, Bus. Horiz. 60 (2017) 677–688, https://doi.org/
[19] B. Reid, Overall Equipment Effectiveness for Additive Manufacturing, James Worth 10.1016/j.bushor.2017.05.011.
Bagley College of Engineering, 2019. 〈https://ir.library.msstate.edu/handle/1166 [46] R.H. Myers, D.C. Montgomery, C. Anderson-Cook. 8.4.5 The Face-Centered Cube in
8/16445〉. Cuboidal Regions, in: Response Surf. Methodol. Process Prod. Optim. Using Des.
[20] M. Baumers, M. Holweg, On the economics of additive manufacturing: Exp., Fourth edition.,, Wiley, Hoboken, New Jersey, 2016, pp. 408–411.
Experimental findings, J. Oper. Manag. 65 (2019) 794–809, https://doi.org/ [47] R.L. Mason, R.F. Gunst, J.L. Hess. 17.3.1. Central Composite Design, in: Stat. Des.
10.1002/joom.1053. Anal. Exp. - Appl. Eng. Sci, 2nd ed..,, John Wiley & Sons, 2003, pp. 583–585. 〈https
[21] J.A. Gopsill, B.J. Hicks, Investigating the effect of scale and scheduling strategies ://app.knovel.com/hotlink/pdf/id:kt003Y2242/statistical-design-analysis/central
on the productivity of 3D managed print services, Proc. Inst. Mech. Eng. Part B J. -composite-design〉.
Eng. Manuf 232 (2018) 1753–1766, https://doi.org/10.1177/ [48] H.R. Lindman, Chapter 7: Higher Way Designs, in: Anal. Var. Exp. Des, Springer
0954405417708217. New York, New York, NY, 1992, pp. 159–189, https://doi.org/10.1007/978-1-
[22] A.J. De Ron, J.E. Rooda, OEE and equipment effectiveness: an evaluation, Int. J. 4613-9722-9.
Prod. Res. 44 (2006) 4987–5003, https://doi.org/10.1080/00207540600573402. [49] M. Jahangirian, T. Eldabi, A. Naseer, L.K. Stergioulas, T. Young, Simulation in
[23] S. Mellor, L. Hao, D. Zhang, Additive manufacturing: A framework for manufacturing and business: A review, Eur. J. Oper. Res 203 (2010) 1–13, https://
implementation, Int. J. Prod. Econ. 149 (2014) 194–201, https://doi.org/10.1016/ doi.org/10.1016/j.ejor.2009.06.004.
j.ijpe.2013.07.008. [50] C.P.T. Hedenstierna, S.M. Disney, J. Holmström, Order book management in 3D
[24] M. Baumers, L. Beltrametti, A. Gasparre, R. Hague, Informing additive printing service operations: A design science approach, in: Trondheim, Norway
manufacturing technology adoption: total cost and the impact of capacity (2016) 10.
utilisation, Int. J. Prod. Res. 55 (2017) 6957–6970, https://doi.org/10.1080/ [51] S.H. Khajavi, M. Baumers, J. Holmström, E. Özcan, J. Atkin, W. Jackson, W. Li, To
00207543.2017.1334978. kit or not to kit: Analysing the value of model-based kitting for additive
[25] G. Costabile, M. Fera, F. Fruggiero, A. Lambiase, D. Pham, Cost models of additive manufacturing, Comput. Ind. 98 (2018) 100–117, https://doi.org/10.1016/j.
manufacturing: A literature review, Int. J. Ind. Eng. Comput. (2017) 263–283, compind.2018.01.022.
https://doi.org/10.5267/j.ijiec.2016.9.001. [52] Loughborough University Additive Manufacturing Research Group, EOS Formiga
[26] S. Nakajima, Introduction to TPM: Total Productive Maintenance, Productivity P100 | Additive Manufacturing Research Group | Loughborough University, 2021.
Press, Cambridge, Mass, 1988. 〈https://www.lboro.ac.uk/research/amrg/amrg-group/facilities/equipment/e
[27] J.A. Garza-Reyes, S. Eldridge, K.D. Barber, H. Soriano-Meier, Overall equipment osformigap100/〉 (accessed 8 April 2021).
effectiveness (OEE) and process capability (PC) measures: A relationship analysis, [53] M. Abdelrahman, T.L. Starr, Quality certification and control of polymer laser
Int. J. Qual. Reliab. Manag. 27 (2010) 48–62, https://doi.org/10.1108/ sintering: layerwise temperature monitoring using thermal imaging, Int. J. Adv.
02656711011009308. Manuf. Technol. (2015), https://doi.org/10.1007/s00170-015-7524-1.
[28] P. De Groote, Maintenance performance analysis: a practical approach, J. Qual. [54] T. Phillips, S. Fish, J. Beaman, Development of an automated laser control system
Maint. Eng 1 (1995) 4–24, https://doi.org/10.1108/13552519510089556. for improving temperature uniformity and controlling component strength in
[29] J.M. Jauregui Becker, J. Borst, A. van der Veen, Improving the overall equipment selective laser sintering, Addit. Manuf. 24 (2018) 316–322, https://doi.org/
effectiveness in high-mix-low-volume manufacturing environments, CIRP Ann 64 10.1016/j.addma.2018.10.016.
(2015) 419–422, https://doi.org/10.1016/j.cirp.2015.04.126.

11
S. Basak et al. Additive Manufacturing 56 (2022) 102904

[55] M. Abdelrahman, T.L. Starr, Feedforward Control for Polymer Laser Sintering [62] R. Oliveira, S.A. Taki, S. Sousa, M.A. Salimi, Global Process Effectiveness: When
Process Using Part Geometry, in: Proc. Solid Free. Fabr. Symp., Austin, Texas, Overall Equipment Effectiveness Meets Adherence to Schedule, Procedia Manuf 38
2015: pp. 492–500. 〈http://sffsymposium.engr.utexas.edu/sites/default/files/20 (2019) 1615–1622, https://doi.org/10.1016/j.promfg.2020.01.123.
15/2015–39-Abdelrahman.pdf〉 (accessed March 1, 2019). [63] J.A. Garza-Reyes, From measuring overall equipment effectiveness (OEE) to overall
[56] S.H. Klippstein, F. Heiny, N. Pashikanti, M. Gessler, H.-J. Schmid, Powder spread resource effectiveness (ORE), J. Qual. Maint. Eng. 21 (2015) 506–527, https://doi.
process monitoring in polymer laser sintering and its influences on part properties, org/10.1108/JQME-03-2014-0014.
JOM 74 (2022) 1149–1157, https://doi.org/10.1007/s11837-021-05042-w. [64] C. Lindemann, U. Jahnke, M. Moi, R. Koch, Analyzing Product Lifecycle Costs for a
[57] E. Westphal, H. Seitz, A machine learning method for defect detection and Better Understanding of Cost Drivers in Additive Manufacturing, in: Proc. 23rd
visualization in selective laser sintering based on convolutional neural networks, Annu. Int. Solid Free. Fabr. Symp. – Addit. Manuf. Conf., Austin, Texas, 2012: pp.
Addit. Manuf 41 (2021), 101965, https://doi.org/10.1016/j.addma.2021.101965. 177–188.
[58] L. Xiao, M. Lu, H. Huang, Detection of powder bed defects in selective laser [65] N. Hopkinson, P. Erasenthiran, High Speed Sintering - Early Research into a New
sintering using convolutional neural network, Int. J. Adv. Manuf. Technol. 107 Rapid Manufacturing Process (2004), https://doi.org/10.26153/TSW/6998.
(2020) 2485–2496, https://doi.org/10.1007/s00170-020-05205-0. [66] HP, HP Multi Jet Fusion technology, HP Development Company L.P., 2017. 〈http
[59] Y. Oh, P. Witherell, Y. Lu, T. Sprock, Nesting and scheduling problems for additive s://cimquest-inc.com/wp-content/uploads/2018/04/HP-MJF-Technology-Techni
manufacturing: A taxonomy and review, Addit. Manuf. 36 (2020), 101492, https:// cal-White-Paper.pdf〉 (accessed 17 January 2020).
doi.org/10.1016/j.addma.2020.101492. [67] Z. Šoškić, G.L. Monti, S. Montanari, M. Monti, M. Cardu, Production cost model of
[60] L.A. Gólcher-Barguil, S.P. Nadeem, J.A. Garza-Reyes, Measuring operational the multi-jet-fusion technology, Proc. Inst. Mech. Eng. Part C J. Mech. Eng. Sci
excellence: an operational excellence profitability (OEP) approach, Prod. Plan, (2019), https://doi.org/10.1177/0954406219837300.
Control 30 (2019) 682–698, https://doi.org/10.1080/09537287.2019.1580784. [68] EOS GmbH, Customer 1zu1 buys EOS P 500 polymer systems, 2021. 〈https://www.
[61] K.M.N. Muthiah, S.H. Huang, Overall throughput effectiveness (OTE) metric for eos.info/en/presscenter/press-releases/eos-p-500–1zu1〉 (accessed 10 November
factory-level performance monitoring and bottleneck detection, Int. J. Prod. Res. 2021).
45 (2007) 4753–4769, https://doi.org/10.1080/00207540600786731.

12

You might also like