You are on page 1of 9

Additive Manufacturing 34 (2020) 101173

Contents lists available at ScienceDirect

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

Design of a laser control system with continuously variable power and its T
application in additive manufacturing
Tim Phillips*, Trevor Ricker, Scott Fish, Joseph Beaman
Department of Mechanical Engineering, The University of Texas at Austin, 204 East Dean Keeton Street, Stop C2200, Austin, TX 78712, USA

A R T I C LE I N FO A B S T R A C T

Keywords: Selective laser sintering (SLS) is one of the most popular industrial polymer additive manufacturing processes
Additive manufacturing with applications in aerospace, biomedical, tooling, prototyping, and beyond. SLS is capable of creating unique,
Laser control functional parts with little waste and no tooling by using a high-powered laser to selectively melt powdered
Process control polymer into desired shapes. This process relies heavily on understanding and controlling the thermodynamics of
Surrogate modeling
the polymer melt process. One of the biggest challenges SLS faces is lack of adequate process control, which leads
Powder bed fusion
to comparatively high component variations. It has been shown that implementing more advanced laser control
techniques enable a higher level of control over the processing temperatures and lead to more uniform com-
ponents. Currently, there are no commercial options for a laser power controller that allows continuously
variable power to be used as a galvanometer system adjusts the laser position. This paper will also discuss the
development and implementation of a galvanometer controller solution that works in conjunction with an off-
the-shelf unit to enable this crucial functionality and will present results showing that, when applied to the SLS
process, a 33 % reduction in thermal variation was achieved.

1. Introduction move the laser between two points with non-constant laser parameters
and at the time of writing there are no commercial laser controllers that
This paper explores the use of a novel laser controller for improving allow for intra-vector power modulation. The lack of this capability is
performance in selective laser sintering (SLS) additive manufacturing. detrimental for both researchers and manufacturers, as laser energy and
The work contained in this paper is a continuation of previous work material processing temperature have been shown to influence
that developed a method of controlling laser power is SLS in order to strength, density, and elongation of components built using SLS [1–5].
improve the consistency of components built [1]. This previous effort In Selective Laser Melting AM of metallic components, energy density
was capable of improving temperature uniformity of SLS components has an even wider influence as it can also impact hardness, micro-
by up to 57 % and strength uniformity by up to 45 %. The im- structure, and wear properties [6–8].
plementation presented in that paper, however, was only applicable to Previous efforts at controlling laser power and, consequently,
simple shapes with constant cross-sections, such as the flexure speci- component properties in AM have revolved around discretizing the
mens presented therein. Since one of the many advantages of additive powder bed into sections wherein laser parameters are constant [1] or
manufacturing (AM) is reduced geometric restrictions [2], this limita- developing completely custom controllers capable of the desired level
tion is unacceptable for a controller intended to be used in an AM of laser control [9,10]. The former approach poses an issue when a
production environment. As such, this paper will detail a new controller component spans multiple of these discretized sections, as introducing
that applies similar laser power control concepts to complex shapes. step changes to laser power in the interior of a component may have
One of the current limitations with laser-based AM is a lack of access detrimental effects. The latter approach, which has been implemented
to process variables within commercial laser controllers. These com- in metal AM, involves recreating much of the functionality already
mercial laser controllers synchronize movement of a galvanometer offered by commercial controllers in order to add the new desired
system with firing of a laser and allow the user to change settings such features and requires the extensive development time, testing, and costs
as speed and power. Most of these controllers, however, require that associated with those tasks. The preferred approach, then, is enable this
each vector be fully defined with constant parameters before execution. crucial intra-vector power control without needing to recreate existing
For example, it is not possible to command the galvanometer system to hardware and software. Enabling this functionality allows for advanced


Corresponding author.

https://doi.org/10.1016/j.addma.2020.101173
Received 11 November 2019; Received in revised form 4 February 2020; Accepted 4 March 2020
Available online 22 April 2020
2214-8604/ © 2020 Elsevier B.V. All rights reserved.
T. Phillips, et al. Additive Manufacturing 34 (2020) 101173

Fig. 1. Example layer scan files and their associated sintering masks for a range of part complexities.

laser control concepts to be implemented in AM machines. Real-time or camera provides 64 × 64 pixels images at a rate of 2243 Hz and has a
model-predictive controllers can be developed to determine optimal moving frame of reference that follows the laser spot. All thermal tests
laser parameters based on current system temperature, model geo- presented in this paper were conducted using ALM PA650, a common
metry, or other factors that influence component quality, and the intra- nylon 12 SLS material.
vector power controller used to create components in the most optimal
way. This paper will present the design of a supplementary controller 2.2. Proposed control technique
that works with an off-the-shelf controller to enable the laser control
tasks discussed and will present results of using this system in an SLS One of the goals of the work presented in this paper is to improve
machine. Dynamic surrogate modeling is used to analyze machine re- temperature uniformity on the powder bed of an SLS machine by
sponse and predict optimal laser powers to take advantage of the new modulating laser power to account for pre-existing temperature gra-
laser control functionality. Since this variable has such a large impact dients. This task is an extension of previous work by the authors that
on component quality, the presented method of increasing the con- showed dynamic surrogate modeling was capable of predicting optimal
trollability of laser power is of significant importance to the AM com- laser powers and that actuating the laser accordingly resulted in sig-
munity. nificant improvements to component uniformity [1]. A shortcoming of
the approach presented in that paper is that it defines a single laser
2. Materials and methods power for each component, meaning large components that span a
significant temperature gradient on the powder surface must either be
2.1. SLS testbed split up or suffer suboptimal performance. This restriction is not ac-
ceptable in AM, where complex components are the norm. Therefore, a
The controller development and experimentation discussed in this more advanced control system is required that does not place any ad-
paper was accomplished on a custom research machine developed at ditional restrictions on component geometry. The system must be able
the University of Texas at Austin. This machine is known as the Laser to collect and analyze thermal data from any arbitrary location on the
Additive Manufacturing Pilot System (LAMPS) and more information powder bed, use that information to identify optimal laser parameters,
on its development can be found in Fish et al. [11]. LAMPS uses an and then actuate the laser in a continuous fashion using the predicted
Iradion Infinity series CO2 laser with a spot size of approximately 575 optimal values.
μm and nominal maximum output of 40 W. The laser output is con- Collection of thermal information is done using LAMPS’ FLIR A6701
trolled with a pulse-width modulated (PWM) signal whose period is MWIR camera. The camera is mounted externally of the LAMPS ma-
selectable by the user, typically in the range of 50–200 μs. Since the chine at an angle of 28° from powder bed normal. To interpret the
input signal to the laser is the variable that is directly being controlled thermal data, the camera images must be corrected for the viewing
by the system developed in this paper, laser energy will be discussed in angle and registered so that the thermal image is in a known coordinate
terms of percent, so that 50 % power is nominally 20 W, and a power frame. The registration process is done by sintering a known geometry
difference of 10 % is the change from 50 % output to 40 % output. and using a least-squares perspective transformation to create a trans-
LAMPS uses a FLIR A6701 mid-wave infrared (MWIR) camera to form matrix that brings the infrared images into machine coordinate
monitor the building process and control powder bed heaters. This space. After an image has been registered, the layer scan files can be
camera provides 640 × 512 pixels images at a rate of 60 Hz. It will also used to identify the sintered regions and extract powder temperatures.
be used in this paper to provide feedback to a dynamic surrogate sin- Some example scan files and their associated sintering masks are given
tering model that utilizes the expanded laser control functionality to in Fig. 1, showing that the process turns difficult to read scan files into
improve consistency of test specimens. A second MWIR camera, a FLIR useable image processing masks for parts of all complexities. An ex-
SC8240, is boresighted with the laser and provides high-speed thermal ample of the temperature extraction process is given in Fig. 2, where
data to monitor how the powder bed responds to the laser energy. This Fig. 2(a) shows the raw thermal image, Fig. 2(b) is the image after it has

2
T. Phillips, et al. Additive Manufacturing 34 (2020) 101173

Fig. 2. Thermal data extraction process showing (a) the raw thermal image, (b) the same image registered into machine coordinate space, (c) the sintering mask, and
(d) the relevant thermal data for sintered regions.

been registered and brought into machine coordinate space, Fig. 2(c) is
the associated sintering mask, and Fig. 2(d) is the relevant thermal
information for the sintered regions.
The thermal data extraction process takes place at the end of each
layer and the information is used to update laser sintering models. Since
there are some positional dependencies on laser fluence due to the
optical track, a unique sintering model is created for each 1 mm x 1 mm
area on the powder bed. The models contain historical laser power and
measured temperature increase data and every time a location is sin-
tered the new data is added to the governing model. A least squares
approach is used with a forgetting factor of 0.8 to identify a relationship
between applied laser power and expected temperature increase at that
given position that tracks with changes in the machine throughout a
build. Before each layer, a pre-sintering thermal measurement is taken
and compared with a desired post-sintering temperature and the re-
levant thermal models are queried to generate predictions of the op-
timal laser power at each position, examples of which are shown in red
in Fig. 3. Once these sintering models are solved, regression is used to
fit a surface of the form given in Eq. 1 to the optimal laser power so- Fig. 3. Power surface showing the surrogate model solutions in red and the
lutions, consolidating them to give a position-dependent laser power power surface in blue.
surface. An example power surface is shown in blue in Fig. 1. As
highlighted earlier, though, no current off-the-shelf controller would 2.3. Design of laser and galvanometer controller
allow for this power surface to be applied. It is necessary, then, to de-
velop additional laser control hardware to execute these optimal set- Originally, the LAMPS machine used a Cambridge Technologies
tings and improve component quality and consistency. EC1000 to control galvanometer positioning and laser firing. As high-
lighted above, this controller lacks the ability to change laser power
P (x , y ) = α 00 + α10 x + α 01 y + α20 x 2 + α11 xy + α 02 y 2 + α30 x 3 + α21 x 2y after a vector command has been issued and is incapable of changing
+ α12 xy 2 + α 03 y 3 (1) the power within a vector. There are, however, many useful functions
in this controller for governing galvanometer motion and it would be a
lengthy, and likely expensive, task to replicate all of its features. The

3
T. Phillips, et al. Additive Manufacturing 34 (2020) 101173

Fig. 4. Signal diagrams for (a) the stock EC1000 controller configuration, and (b) the system with the Position-Based Laser Power controller.

signal diagram for the stock EC1000 controller is presented in Fig. 4(a) the powder surface. The time it takes for the FPGA to solve the surface
and shows that it interprets commands from the host LAMPS computer equation is significantly faster than the selected laser duty cycle period
and sends the corresponding signals to the galvanometer controller and of 100 μs, meaning the commanded laser PWM should closely match
laser, synchronizing their timing to ensure proper laser positioning. The the power surface solution at the current galvanometer position.
approach considered in this paper is to allow the stock controller to
maintain command over the galvanometer system and implement a 2.4. Evaluation of the laser controller
separate, complimentary controller that will take control over the laser
power signal. After implementing the PBLP on a NI cRIO, static and dynamic tests
The so-called Position-Based Laser Power (PBLP) controller is were performed to validate the output of the controller. First, a static
placed between the EC1000 and laser and receives the original laser position calibration was conducted to map the analog voltage feedback
power signal from the EC1000, real-time position feedback from the signal from the galvanometers to Cartesian coordinates on the powder
galvanometer controllers, and a mathematical representation of the bed. This calibration was done by commanding the galvanometer to
desired power surface as inputs. The PBLP controller is housed on a different, known positions across the powder bed and recording the
National Instruments (NI) compactRio (cRIO) system with a Field- signal measured by the AI/O module in the PBLP. The powder bed was
Programmable Gate Array (FPGA), a NI 9205 analog input module, and sampled in 5 mm increments across the entirety of its surface. Spot-
a NI 9402 digital input/output module. The NI 9205 module is used to checking showed that there was no cross talk between the voltage
measure the galvanometer position signal and has an update rate of 4 μs signals from the two galvanometers, so it was not necessary to take
(8 μs to collect both X and Y signals). The 9402 module has an update measurements over the entire position matrix, just over the individual
rate of 55 ns and is used to measure the laser power signal and generate position arrays. While this process does not calibrate the absolute po-
the PWM signal sent to the laser. The new signal diagram for EC1000 sition of the laser on the powder bed, it ensures the PBLP and EC1000
with the PBLP is presented in Fig. 4(b). have the same reference frame. Since the stock EC1000 has full control
The galvanometer position signal acquired by the PBLP controller over galvanometer position, adding the PBLP to the laser control system
comes from the capacitive feedback on the galvanometers that is used will not have any negative impact on the absolute positioning of the
in the galvanometer position controller. The signal is continuously laser.
monitored and used to solve a mathematical power surface to de- After position calibration, dynamic testing was performed by scan-
termine what the desired power is at the current laser position. The ning the laser across the powder bed at different speeds while applying
original laser power signal generated by the EC1000 is used as a trigger a known power map to the PBLP. The actual output of the laser control
for the PBLP controller, which analyzes the signal to determine if the system was compared with the power defined by the power surface.
laser is supposed to be firing or not firing at any given moment. This These dynamic tests consisted of 50 mm vectors that linearly increased
signal is monitored by the FPGA every microsecond and immediately the desired laser power from 10 % at the beginning to 90 % at the end
triggers the PWM output to the laser when the original signal is high, of each vector. Laser speed was varied from 500 mm/s to 5500 mm/s in
causing a maximum laser on delay of one microsecond. When the ori- 1000 mm/s increments. Galvanometer position feedback and laser
ginal laser power signal indicates the laser should be off, the PBLP sends power output by the PBLP were recorded every 89 μsec.
a zero signal on the next PWM period, causing a maximum off delay of A second dynamic test was performed by driving LAMPS’ 40 W CO2
one period, depending on when the laser off signal was received in the laser with the PBLP and directing the galvanometers onto a piece of
PBLP PWM cycle. This signal structure allows the PBLP controller to Mylar as it scans a test pattern while varying the laser power. Mylar is
have no knowledge of the scan path, yet still actuate the laser at the commonly used in laser calibration as it changes from black to varying
correct time intervals and allows the EC1000 to maintain full control levels of gray as it is exposed to higher laser energy levels and was used
over positioning and timing of the system. In the testing presented here to give a visual indication of how the laser power was being
herein, a 3rd order surface is used to describe desired laser power across modulated. The test pattern, shown in Fig. 5(a), consists of continuous

4
T. Phillips, et al. Additive Manufacturing 34 (2020) 101173

Fig. 5. PBLP test showing the (a) laser scan and (b) laser power surface.

lines oriented horizontally, perpendicularly, +45°, and −45° spaced 5 3. Results and discussion
mm apart over the span of a 40 mm x 40 mm rectangle. Prior testing
showed that 30 % laser power is the threshold for marking this Mylar 3.1. PBLP bench testing
material and powers above 46 % were visibly indistinguishable.
Therefore, laser power was constrained in this test to be between those Static testing of the galvanometer position feedback signals revealed
two values so that results could be observed with a naked eye. The that the voltage signal obtained is highly linear with the commanded
power surface started at 46 % power in the upper right corner at x = 20 position, allowing a simple linear equation to fully define their re-
mm, y = 20 mm and varied linearly to 30 % power in the lower left lationship. The data and fit statistics are given in Fig. 7.
corner at x= −20 mm, y= −20 mm, as depicted in Fig. 5(b). Dynamic bench testing of the PBLP system showed that there was a
small amount of error that scaled with the scanning speed of the gal-
vanometers, where the power output lagged behind the value defined
2.5. Evaluation of laser controller in selective laser sintering by the power surface. Fig. 8 presents actual laser output as a function of
position, where the black dashed line is the profile defined by the power
A potential software solution to overcome lack of intra-vector power surface. The galvanometer was scanning in the positive direction during
control is to subdivide each vector into discrete sections with constant this trial so that the position and power output are increasing with time.
power. This approach would allow the power to be changed while The solid colored lines show the power signal generated by the PBLP as
scanning a vector on the interior of a component without needing ad- the galvanometer scans the vector at different speeds. As the galvan-
ditional hardware, such as the PBLP presented. However, this technique ometer speed increases, more deviation from the ideal power profile is
has its drawbacks, as each power change instance will require the observed. The amount of error is relatively constant for each galvan-
galvanometers to come to a physical stop before the new power setting ometer speed regardless of commanded power, indicating the error is
can be applied. The stop event is not caused by a limitation in the caused by delay in acquiring galvanometer position and solving the
galvanometer hardware, but is a constraint imposed by the currently power profile. The error is quantified in Fig. 9, where it is plotted
available galvanometer control software. Implications of this approach against the rate of change in the commanded power output defined as:
are explored by using LAMPS to evaluate the temperature increase
∂P dx ∂P dy
along a series of vectors with varying power section granularity. Testing ∇P = +
∂x dt ∂y dt
consisted of scanning 50 mm vectors while varying the laser power
from 10 % to 90 % of the maximum laser output using power steps of 10 Preliminary investigation of the PBLP system inside a SLS machine
%, 5 %, and 1 % of the laser output. The FLIR SC8240 high speed MWIR showed that the laser was capable of intra-vector power modulation
camera was used to record the temperature on the powder bed and when scanned across a piece of Mylar in the test pattern described in
thermal results from this testing were compared with those produced by Fig. 5. The resulting Mylar sheet is seen in Fig. 10, where it is clear that
the PBLP given the same control task. the intensity of the vectors vary in a way consistent with the power
Further testing of the PBLP system was performed by utilizing the pattern in Fig. 5(b). No considerations were made in the test pattern
laser control technique described in Section 2.2 to complete an entire where the laser vectors intersect, resulting in more energy deposited in
SLS build and comparing temperature uniformity results with a similar those areas that cause the bright spots on the Mylar sheet. This initial
build performed using a constant laser power. During the PBLP build, test indicates that the PBLP controller works as expected and it is ap-
the coefficients of the power surface predicted by the surrogate models propriate to move on to the next tests to quantify the effects of the PBLP
are sent to the PBLP controller and used to dynamically vary the laser system and its effect on powder temperature.
power. All other machine parameters were kept constant between the
two builds, which consisted of 49 identical components each. The
3.2. Results of selective laser sintering implementation
components were shaped like miniature I-beams with a 10 mm x 10 mm
footprint with 10 mm height. The full dimensions and arrangement can
Typical laser scanning speeds in the LAMPS SLS machine are on the
be seen in Fig. 6. Thermal image data was recorded throughout the
order of 1500 mm/s, making the error introduced by the PBLP system
entirety of the build process using the FLIR A6701 MWIR camera and
on the lower end of the numbers presented in Fig. 9. Additionally, the
was used to compare the two control techniques.
period of the laser PWM signal was chosen to be 100 μs with resolution
of 1 μs, meaning laser power is only capable of changing in 1 % in-
crements and any error within 0.5 % of the desired value is rounded
out.

5
T. Phillips, et al. Additive Manufacturing 34 (2020) 101173

Fig. 6. Dimensions of (a) SLS test part and (b) their arrangement within the build chamber, dimensions in mm.

Fig. 9. PBLP power output error.


Fig. 7. Static galvanometer position calibration.

Comparing results of the PBLP controller against those obtained by


the software-only solution of sectioning vectors showed multiple ad-
vantages when using the new controller. Fig. 11 presents the tem-
perature increase measured on the powder surface for different sub-
section granularities, where the solid blue line is the measured
temperature, the dashed blue line is a linear fit to the data, and the
expected temperature range is highlighted in yellow. As expected,
coarse steps in power result in a noticeable stair-stepping effect rather
than the intended linear increase. At small subsection sizes, such as 1%
increments presented in Fig. 11(c), the stair-stepping is no longer ap-
parent, but the temperature increase does not match well with the
predicted values. At this setting, the temperature increase was incon-
sistent between trials, resulting in the large deviations from the linear
fit seen in Fig. 11(c). In addition, the temperatures measured are higher
than expected due to the accumulation of end-of-vector effects where
the galvanometer is accelerating or decelerating and inputting more
energy at the subsection interfaces. The presence of these acceleration
regions is also evidenced by the fact that the 1 % incremented vector
took 25 % longer than the same vector without sectioning. The PBLP
single vector results, seen in Fig. 11(d), show good agreement with the
Fig. 8. Power as a function of position for different galvanometer speeds. predicted temperature increase and did not slow down the sintering

6
T. Phillips, et al. Additive Manufacturing 34 (2020) 101173

layer build showed drastic improvements in temperature uniformity


when compared to an open-loop build. The IR images collected during
each layer were analyzed to determine the post-sintering temperature
within each component by averaging the relevant IR image pixels. The
resulting layer-wise post-sintering temperatures of both builds for
components at the corners and center of the I-beam tests are given in
Fig. 12. To compare the proposed control system with the industry-
standard fixed-power build strategy, range and standard deviation were
computed on each layer for the two build strategies using the layer-wise
component temperatures. Once the layer-wise range and standard de-
viation values were calculated, a comparison could be made by de-
termining the percent change in those values on each layer when using
the proposed control strategy. Fig. 13 shows the percent reduction in
layer-wise temperature range and standard deviation among all 49
components when using the proposed control system compared with
the same build that used a fixed laser power. There are some undesir-
able results during the first 4 layers of the build, where the thermal
results show both positive and negative effects of the controller. The
likely source of these oscillations is initialization of the surrogate
models, which were initialized with only a single data point indicating
the lower laser threshold. The first layer of the PBLP build used a pre-
determined fixed laser power to provide a second data point in the
sintering models and at layer two, the feed-forward surrogate model
approach took over in predicting the optimal laser powers. If the in-
Fig. 10. Results of Mylar PBLP testing. itialization stage were extended so that a few more layers were sintered
with the pre-determined laser power, the results would likely be im-
process. This controller, then, is superior to a software-only solution in proved when the laser power controller took over due to the increase in
terms of accuracy, repeatability, and speed. thermal data available when making the laser power predictions. An-
Evaluation of the PBLP and feed-forward control system for a 100- other dip in performance is found on layer 83 that caused the standard

Fig. 11. Temperature increase comparison of different laser control strategies, showing the idealized result in yellow and results when using (a) 10 % increments, (b)
5 % increments, (c) 1 % increments, and (d) The Position-Based Laser Power controller.

7
T. Phillips, et al. Additive Manufacturing 34 (2020) 101173

Fig. 12. Results of I-Beam build showing the layer-wise post-sintering temperature for 5 component locations on (a) the baseline fixed-power build and (b) the laser
power controlled build.

error when used under conditions typical to Selective Laser Sintering


and was faster and more accurate than a software-only solution.
The PBLP controller was combined with a previously developed
feed-forward laser power control strategy and installed on a research
SLS machine. Evaluation of the laser control system showed drastic
improvements over open-loop control systems, where standard devia-
tion of component temperatures was reduced by 33 %. Previous studies
of flexure specimens showed that this type of improvement in tem-
perature uniformity yields similar improvements in component strength
uniformity [1]. The addition of the PBLP controller allows the feed-
forward technique to be applied to more complex components, im-
Fig. 13. Improvement in temperature uniformity when using the PBLP and
proving the likelihood of creating consistently high-quality parts. The
feed-forward control system.
self-contained nature of the controller presented should allow for ret-
rofitting commercial machines with the laser control features pre-
deviation of post-sintering temperatures to be worse than that of the sented, increasing their performance.
baseline test, though the range was still improved. It is unclear what A few avenues of future work have been identified that may im-
caused this temperature drop, but the control system self-corrected and prove the performance of the control system presented. Since the sur-
quickly brought the post-sintering temperature standard deviation back rogate models track which areas of the powder bed are sintered on each
within the desired range. layer, it may be possible for the system to identify when an overhang is
Layer 70 presents the transition from the skinny section of the I- being created and adjust the laser power to account for the different
beam back to the full section. This step geometry change creates un- build dynamics in that location to improve temperature uniformity.
supported overhangs in the components where material is now being Decreasing the thermal fluctuations on the first few layers of a build is
lased above un-sintered powder. These features typically cause concern also important and may be as simple as pre-initializing the thermal
in SLS systems because the thermodynamics of the sintering process are models prior to building any critical components. Lastly, it may be
different when lasing on top of powder vs previously melted material. possible to improve the way desired post-sintering temperature is se-
The thermal data in Fig. 12 show that both the fixed power and PBLP lected. Currently, design of experiments testing and destructive mea-
builds have a sharp drop in temperature at layer 70 due to these surements are used to empirically find this value, but a model may
overhangs. The temperatures experienced by the PBLP build, however, provide superior results. It is also possible that such a model would
are closer to the desired building temperature (468 K, in this case) than specify a non-constant desired temperature dependent on component
those experienced by the fixed power test. Additionally, the percent geometry, which could be incorporated into the PBLP system. Overall,
improvement in temperature uniformity presented in Fig. 13 shows that the controller presented in this paper shows promise at improving the
at layer 70, the range and standard deviation of temperatures are ap- quality of components built using SLS. As poor repeatability is a com-
proximately 60 % smaller during the PBLP build, indicating the con- monly cited downside of additive manufacturing, this control system
troller presented is superior for building these features. Taking the has the potential to improve the utility and confidence in SLS and en-
average of the reduction in post-sintering temperature range and able new capabilities for researchers and manufacturers alike.
standard deviation across all layers yields overall performance metrics
of 52 % and 33 % improvement, showing the laser power controller
creates significantly more uniform components. CRediT authorship contribution statement

Tim Phillips: Writing - original draft, Writing - review & editing,


4. Conclusions Software, Validation, Investigation, Conceptualization. Trevor Ricker:
Conceptualization, Methodology. Scott Fish: Supervision. Joseph
This paper presented a novel solution to enable intra-vector power Beaman: Supervision, Funding acquisition.
modulation capabilities using off-the-shelf components. A supplemen-
tary controller, called the Position-Based Laser Power controller, in-
tercepts the laser power signal generated by a commercial laser and Declaration of Competing Interest
galvanometer control system and modifies the signal before passing it
to the laser. The laser signal can be modified according to a parametric The authors declare that they have no known competing financial
surface so that the new laser power signal is a function of galvanometer interests or personal relationships that could have appeared to influ-
position. Benchtop testing showed this technique had acceptably small ence the work reported in this paper.

8
T. Phillips, et al. Additive Manufacturing 34 (2020) 101173

Acknowledgments [5] S. Singh, V.S. Sharma, A. Sachdeva, S.K. Sinha, Optimization and analysis of me-
chanical properties for selective laser sintered polyamide parts, Mater & Manuf
Process 28 (2013) 163–172.
The authors would like to acknowledge the instrumental research [6] Q. Jia, D. Gu, Selective laser melting additive manufacturing of Inconel 718 su-
support of the US Department of Defense (DoD)Air Force Research peralloy parts: densification, microstructure and properties, J. Alloys. Compd. 585
Laboratory (AFRL)Project # FA8650-17-C5716 P00003: Laser Additive (2014) 713–721.
[7] H.-Z. Jiang, Z.-Y. Li, T. Feng, P.-Y. Wu, Q.-S. Chen, Y.-L. Feng, S.-W. Li, H. Gao, H.-
Manufacturing Pilot Scale (LAMPS) III: Advanced Process Monitoring J. Xu, Factor analysis of selective laser melting process parameters with normalised
and Control (PI: Scott Fish). quantities and Taguchi method, Opt. Laser Technol. 119 (2019).
[8] K. Kempen, L. Thijs, J.V. Humbeeck, J. Kruth, Processing AlSi10Mg by selective
laser melting: parameter optimisation and material characterisation, Mater. Sci.
References Technol. 31 (8) (2015) 917–923.
[9] B. Lane, S. Mekhontsev, S. Grantham, M.L. Vlasea, J. Whiting, H. Yeung, J. Fox,
[1] T. Phillips, S. Fish, J. Beaman, Development of an automated laser control system C. Zarobila, J. Neira, M. McGlauflin, L. Hanssen, S. Moylan, A. Donmez, J. Rice,
for improving temperature uniformity and controlling component strength in se- Design, Developments, and Results From the NIST Additive Manufacturing
lective laser sintering, Addit. Manuf. 24 (2018) 316–322. Metrology Testbed (AMMT)," in Solid Freeform Fabrication Symposium, Austin,
[2] C. Weller, R. Kleer, F.T. Piller, Economic implications of 3D printing: market (2016).
structure models in light of additive manufacturing revisited, Int. J. Prod. Econ. 164 [10] B. Fulcher, Real-time Pyrometer Feedback and Control in Metal Powder Bed Fusion,
(2015) 43–56. Austin: Paper Presented to Solid Freeform Fabrication Symposium, (2018).
[3] B. Caulfield, P.E. McHugh, S. Lohfeld, Dependence of mechanical properties of [11] S. Fish, J.C. Booth, S.T. Kubiak, W.W. Wroe, A.D. Bryant, D.R. Moser, J.J. Beaman,
polyamide components on build prameters in the SLS process, J. Mater. Process. Design and subsystem development of a high temperature selective laser sintering
Technol. 182 (2007) 477–488. machine for enhanced process monitoring and control, Addit Manuf 15 (2015)
[4] I. Gibson, S. Dongping, Material properties and fabrication parameters in selective 60–67.
laser sintering process, Rapid Prototyp. J. 3 (4) (1997) 129–136.

You might also like