Solid Freeform Fabrication 2019: Proceedings of the 30th Annual International

Solid Freeform Fabrication Symposium – An Additive Manufacturing Conference

Reviewed Paper

AN OPEN-ARCHITECTURE MULTI-LASER RESEARCH PLATFORM FOR

ACCELERATION OF LARGE-SCALE ADDITIVE MANUFACTURING (ALSAM)

William Carter1, Michael Tucker1, Michael Mahony1, David Toledano1, Robert Butler1,
Subhrajit Roychowdhury1, Abdalla R. Nassar2, David J. Corbin2,
Mark D. Benedict3, Adam S. Hicks3
1
GE Research, 2Penn State, 3United States Air Force

Abstract
As Selective Laser Melting (SLM) technology matures, researchers and engineers
responsible for transitioning the technology from rapid prototyping into manufacturing are gaining
a better understanding of the opportunities with this revolutionary technology. A step for
accelerating solutions is to allow researchers complete access to all aspects of the process for
experimentation. As part of an AFRL-sponsored program with America Makes, a production-
grade SLM machine (a Concept Laser M2) will be enhanced to allow operation with either the
original OEM controls and scan path generation or an open-source set of software developed under
America Makes programs. This machine will be referred to as the ALSAM Platform and will be
delivered to the Air Force along with the source code for the open scan path generation software
(written in C++) and the open machine controller (written in LabVIEW™).

Introduction
Interest in the production of large-scale parts using Selective Laser Melting (SLM) has
grown as additive manufacturing has matured and, to respond to this interest, machines are
entering the market with increased build volumes. Single laser systems can successfully print large
components for prototypes or limited production but build times can become untenable for mass
production. A natural solution to allow larger parts on economical time scales is to use multiple
lasers. Proprietary multi-laser solutions are emerging, but to date there is no consensus regarding
optimal strategies for employing multiple lasers and issues including stitching, alignment, and
coordination arise.
To address these concerns, the US Air Force sponsored a series of four research programs
through America Makes (America Makes is managed and operated by the National Center for
Defense Manufacturing and Machining, NCDMM). These programs, summarized in Table 1, were
aimed at developing open-source and open-architecture SLM for metals to foster creative research
in the field. Three of these programs are complete, and a fourth is ongoing. This paper will give
an overview of these programs.

13
Table 1. America Makes “Open SLM” Projects. America Makes sponsored four research projects
related to open-source, open-architecture additive manufacturing systems.

Project #1:
America Makes Project 4039
Title: Open Source Process Control for Powder Bed Additive Manufacturing Research
Status: Complete
Participants: GE Research, GE Aviation, Lawrence Livermore National Laboratory
Dates: March 2015 – August 2016
Summary: Open-source computer programs for laser-based powder bed fusions systems
were written for both scan path generation and machine control. The resulting
software was implemented on multiple SLM machines.
Project #2:
America Makes Project 4040
Title: An Open, Layered Protocol for Powder Bed Additive Manufacturing for
Synchronizing Heterogeneous Sensors and High-Speed Data Acquisition
Status: Complete
Participants: Applied Research Lab at Penn State, Honeywell International Corporation,
Northrop Grumman Corporation, 3D Systems
Dates: February 2015 – February 2017
Summary: A standard protocol was developed and demonstrated for specifying the
behavior of SLM systems, including monitoring and control.
Project #3:
America Makes Project 4051
Title: A Flexible Adaptive Open Architecture to Enable a Robust Third-Party
Ecosystem for Metal Powder Bed Fusion Additive Manufacturing Systems
Status: Complete
Participants: GE Research, Rensselaer, MatterFab, Applied Research Lab at Penn State,
Dates: February 2016 – July 2018
Summary: An open-architecture control system for laser-based powder bed fusions systems
was designed and demonstrated on two SLM machines using the software from
the prior program.
Project #4:
America Makes Project 3024
Title: Acceleration of Large Area Additive Manufacturing (ALSAM)
Status: Ongoing
Participants: GE Research, Applied Research Lab at Penn State
Dates: Dec. 2018 – March 2022
Summary: A commercially available 2-laser Concept Laser M2 machine will be modified
to allow operation using either the OEM-supplied scan path generation and
controls software or the America Makes open-source software. This system will
be delivered to the Air Force, including the source code for all software, by the
end of 2020.

14
 Project #1: America Makes #4039
Open Source Process Control for Powder Bed Additive Manufacturing Research
March 2015 – August 2016

The objective of program was to develop protocol and open-source programs that would
take any STL file as input and generate a machine-independent protocol for manufacturing the part
on an SLM system. This protocol is termed the SCAN protocol. Another protocol, termed the
LAYER protocol, was implemented as an intermediate step between the STL file and SCAN file.
Both file protocols conform to the eXtensible Markup Language (XML) format. Demonstration
software was written to generate both LAYER and SCAN files from STL files.
The sequence is illustrated in Figure 1, where computer programs are indicated as yellow
boxes and protocols are indicated as blue boxes. The STereoLithography (STL) format and the
Additive Manufacturing File (AMF) format are established file formats for representing geometric
objects. At least one of these files is required as input for the LAYER generator. The LAYER
protocol describes the physical attributes of all contours and regions comprising a specific layer.
The SCAN protocol describes the laser scan parameters to fuse the layers in an SLM machine.
Both protocols are independent of SLM machine type, with the goal of making them
comprehensive, simple, scalable, and extensible.

Machine Profile

Layer Scan
Generator Generator Machine
Slices the section curves
S li ces the section curves
Controller
Inputs:
Triangles Loyer file
2 thickness Stub for user-supplied
Start/End Z va lu es proprietary process profile
Region Input from Input - long XML
STL file names string.
Outputs Output-long XML
Po ly li nesfor string
internal and Hatching scheme
external contours Custom Power/Speed
Regions (how algorithms
defined?) Outputs
Z heights, contours, hatches
Machine agnostic build file
Environment requirement
Temper-atures
STL or AMF Layer Melt pool specs? Scan
File
Points
Triangles
Colors
Finishes
Materials
Supports
Etc.

Figure 1. Scan Path Generation Sequence. Yellow boxes indicate computer programs and blue boxes
indicate file formats.

15
 LAYER Protocol
The LAYER protocol was designed during the project to be independent of SLM machine
type and describes the physical attributes of a given layer. Multiple layers form a part and each
LAYER is defined with a specified thickness. Regions can be tagged with feature types. For
example, contours can be tagged with the part name and serial number. Additional tags are used
to define regions of support material, bulk material, sidewalls, upward- or downward-facing
surfaces, etc. Anticipating future improvements to SLM, regions may be tagged with differing
materials. All the details within the LAYER protocol are in standard XML text files that can easily
be inspected or modified. The protocol is illustrated in Figure 2.
Details of the LAYER schema are described in other publications by the authors [1]. In the
context of the XML file, Types have elements and attributes. The elements in the LAYER XML
file are described in this section. The elements in the LAYER file schema include VertexType,
VertexListType, EdgeType, RegionType, SliceType, and LayerType.

THICKNESS

COORDINATE SYSTEM
LAYER

VERTEX LIST Vl
VERTEX 15--~ V2
~--~

SLICE TYPE

REGION TAG(S)

EDGE START

END

Figure 2. The LAYER Protocol Hierarchy.

SCAN Protocol
The SCAN protocol was designed during the project to be independent of SLM machine
type and is used to direct a Machine Controller. The file describes the laser position, speed, and
directions, and scan parameters for each layer. Each beam path represents one continuous
translation of the beam relative to the powder bed and consists of a set of parameters that are
indexed in time. The parameters represent any adjustable property of the SLM process such as
coordinates, laser power, beam size, etc. Additionally, they can contain diagnostic cues for

16
information that needs to be recorded at specific points in time such as camera image, stage
position, or temperature. The SCAN file is stored in the XML (eXtensible Markup Language)
format because the file is intended to be both human and machine readable and the visualizers and
editors for this format may be readily developed. The team developed the detailed technical
requirements for the SCAN protocol with the understanding that a successful protocol will
continue to evolve long after the completion of the program. The content of the SCAN file is
illustrated in Figure 3.
An interesting benefit of the machine-independent approach taken here is that, in addition
to providing a standard seamless interface to SLM machines, the protocol can be used to set
boundary conditions for finite element analysis (FEA). This will involve communicating the beam
paths to finite element software through a file format or software automation interface, for instance
ActiveX/COM. This advantage has already been exploited by a NASA-funded team at Rensselaer
[2]. By leveraging these simulations, optimal beam paths may be developed over time.

TRAVELER ID

BUILD SYNC DELAY

TRAJECTORY TAG
TYPE
PATH
~ - - - ~~---------=- SEGMENT ID
SEGMENT
POWER

VERTEX
VELOCITY PROFILE LIST

VELOCITY PROFILE

Figure 3. SCAN data and geometry hierarchy

Further details of the SCAN schema are described in other publications by the authors [1].
In the XML schema, types have elements and attributes. The elements used in the SCAN XML
file are described in this section. The elements used in the SCAN file schema include BuildType,
VelocityProfile, ListType, VelocityProfileType, TrajectoryType, PathType, Vertex, and
SegmentType

17
 LAYER and SCAN Generator Programs
A program written in C++, called LAYER Generator, was developed to demonstrate the
conversion of STL files into a series of slice layer descriptions per the file format of the LAYER
protocol. The slicing of the STL file itself is a well-established capability and it was not our
intention to reinvent these algorithms. An open-source program, Slic3r, was used to perform this
task during execution of the layer generator. Slic3r is free software developed by Ranellucci [3]
with help of contributors in the open source community. This program is primarily aimed at fused
deposition modeling; only the slicing portion of the code is used here. The required executables
can be downloaded from the web. No source code of Slic3r is required, though available in the
open community through GitHub.com.
The GE Global Research team developed a C++ program, called SCAN Generator, to
demonstrate the conversion of LAYER files into SCAN files. The program is a command-line
driven executable that allows an organization to store a library of laser velocity profiles and assign
the velocity profiles to regions of a specific type defined by a tag. Scalable Vector Graphics (SVG)
files can be generated for scan path visualization using conventional programs, or they may be
viewed in the LabVIEW™-based Machine Controller (in virtual machine mode) discussed below.

Project #2: America Makes #4040

An Open, Layered Protocol for Powder Bed Additive Manufacturing for Synchronizing
Heterogeneous Sensors and High-Speed Data Acquisition
February 2015 – February 2017

Under this America Makes program, referred to as Open Protocol program for short,
communication methods were developed and demonstrated for SLM systems. This work was led
by the Applied Research Laboratory at Pennsylvania State University in collaboration with
Honeywell International Inc., Northrop Grumman Corporation and 3D Systems, Inc. The
developed hybrid open protocol system consisted of three components: open transmission of build-
plan data, low-speed (10 Hz) communication of machine conditions, and high-speed (100 kHz)
communication of laser beam position and status.
The developed protocol enabled specification and extraction of scan path and
communication between a SLM system and heterogeneous systems for condition monitoring, data
acquisitions, and in-process sensing. In particular, the developed methods provided
• access to critical data required for process modeling and optimization,
• synchronization of the process state with sensors for in-process monitoring, and
• monitoring and recording of machine condition and status.

Under the original America Makes program, all components of the open protocol system
were demonstrated on a commercial 3D Systems ProX-200 SLM machine. Later, the framework

18
was extended to other SLM systems including 3D Systems ProX-320, EOS M280, and Concept
Laser M2 systems.

High-speed and Low-speed protocols

Two separate methods were utilized for transmission of real-time process data at 100 kHz
and machine monitoring data at 1-10 Hz, respectively. High-speed transfer was carried out using
a protocol similar to XY2-100 protocol to provide real-time, synchronous, digital transmission.
Lower speed measurements are better suited for transmission using an Ethernet-based protocol,
where they may be monitored remotely by networked systems. Thus, an MTConnect-like protocol
was adapted for transfer of machine and management-level data. An illustration of the physical
protocol implementations is provided in Figure 4.

Laser Trigger

Laser Controller
Laser Set Power External
J: Monitoring
oo·
::;-
Laser X Position system
- V>
"C
m
High Speed
Laser Scanner Card Laser Y Position Protocol
- m
a.
V>
- - -0
<U -0
oo·
::i
<U
a.
Vl
u
0
..,
Q)
ii, :: e
0
_, 0.
Measured Laser Power
ProX200 Electronics
Z Defocusine Position
Real-Time Data
i Computer with - Recording

..,... ..
l!l
,.3' ~ g ~
embedded FPGA Host PC
::i
~
'O
~ .'?
n
C,
~
'O
a [ ,.
0
:,

..,.
'2 C

... ~g
;.
:,.
;;
i:r
Q
;i

I::
~

Figure 4. Schematic layout of the high and low-speed measured signals

During the SLM process, both high-speed and low-speed data are buffered and saved to
disk. Network-shared variables are utilized to synchronize the high-speed and low-speed data
streams via a sampling loop executed at 10 Hz. This enables the extension of the system across a
network with separate computers capturing high-speed and low speed data. All data are saved to
separate files in a local or networked location.
High-speed data are made available to heterogeneous sensors interfaced to a real-time data
acquisition (DAQ) system while low-speed data are available to sensors interfaced with a
Windows-based computer (PC). Sensor outputs (typically analog voltages) can be connected to
available inputs on the real-time DAQ system in parallel with high-speed data. Both sensor and

19
high-speed data are buffered and transmitted to the host PC for storage. For additional details, see
[4].
Sensors operating at or less than 10 Hz or which require a windows-based interface are
synchronized to the high-speed measurements on a PC. Published low-speed data along with
sampled high-speed data are synchronized using network shared variables. The low-speed protocol
is independent of the update rate of attached sensor. However, low-speed sensors are continuously
sampled at a fixed (10 Hz) rate on the on the host PC. Data collected across a layer is saved locally
or to a networked location to a separate file.

Implementation, Extensions and Technology Transitions

Under the Open Protocol program r, the protocol was demonstrated on a 3DSystems ProX-200
machine. Two sensing systems were also interfaced with the high-speed and low-speed protocols:
An Ocean Optics HR2000+ES spectrometer and an ARL Penn State proprietary multi-spectral
sensor [5]. Because all data streams were indexed and synchronized, sensor data could be easily
overlapped for later comparison with post-process computed tomography scans (Figure 5).

e
Ifg
-30
..
~
,... .,
~
.JS

,-----------
,
(
J
~
0. -40 ~ - - -- - - - -- - - -- -
"O
3
co
>- -45
~............,..~ -
(
J
-40 -35 .30 ·25 ·20 -·· ..J
X 5cann@r Position (mm) X Build Plate Position (mm)

Figure 5. Results of the ARL PSU Open Protocol System. Scanner position captured by the system
(Left). Multi-spectral sensor data layer with potential flaws identified (Middle). Computed tomography
scan showing lack of fusion defects in same build layer (Right).

Following the completion of the America Makes program, a follow-on effort funded by
Sandia National Laboratory and in collaboration with 3D Systems, Inc. enabled transition and
installation of the Open Protocol system on a ProX-200 system at Sandia National Laboratory. The
system included the ARL Penn State multi-spectral sensor and an optical emission spectrometer.
The system is still operational. An illustration of the sensor layout and a GUI used for operation
of the Open Protocol system is provided in Figure 6.
A parallel effort, funded by DARPA also enabled extension of the Open Protocol System
to a wider class of SLM systems, including the 3D Systems ProX-320, EOS M280, and Concept
Laser M2 system. All efforts utilized the implementation framework developed under original

20
America Makes funded effort, though these later efforts focused primarily on data acquisition from
high-speed sensors at a reduced rate of 50 kHz. These efforts also captured commanded, rather
than actual scanner positions to reduce the need for post-process data filtering.

StMting la't!r umbe-r

StartC.pt11<:.;•' - - - - - - - - StoP ~ ptwe

Curre-nt spKtr~t!f
me.ilSUr!fflf:nl

Figure 6. Layout of Sensors installed (Left) and simple GUI developed for open protocol system for use
by Sandia National Laboratory (Right).

Project #3: America Makes #4051

A Flexible Adaptive Open Architecture to Enable a Robust Third-Party Ecosystem for
Metal Powder Bed Fusion Additive Manufacturing Systems
February 2016 – July 2018

Under this 28-month research program completed in 2018, GE and Rensselaer formed a
team that developed, demonstrated, and documented a fully open-architecture control system for
SLM. The control system includes subsystems that were tested and integrated into a research SLM
machine at Rensselaer’s Center for Automation Technologies and Systems [6]. Demonstration
parts were generated to show the full applicability of the open-source software, working on the
open-architecture control system, with full control of laser scan paths to generate SLM parts.

Hardware Elements
The SLM process must be carried out in an inert atmosphere under controlled temperature
and pressure. Further, SLM systems must contain safety interconnects, alarms and lighting as basic
features. The fundamental hardware subsystems of a SLM machine (shown in Figure 7) are listed
below:
1. Optical System consisting of laser, scanning system, and focusing system
2. Motion Control consisting of build platform motion control, recoater motion control, and
powder hopper motion control
3. Environment Control consisting of gas flow control (O2 / N2 / Ar / Air) and water cooling
4. Safety System consisting of door lock and interconnect, alarm and Lights

21
 Control Elements

Machine -
Status

-. ...__1 D

_. 6
Laser

E-Stop

Argon

Nitrogen

Figure 7. Schematic of a Typical SLM Machine.

Z,Yaw, Pitch, Roll stage

Laser Control & Status !Servo Control)

Laser Control & Status

2 • Re-Coater 1servo Controll

Scanner Control & Status

Focus Control

■ OlCHCl

~
\
:'l,i'

Air
A00£D f[ATURES Handling
3 MORE SERVOS IBUllD PlATrORM)
ADO(O IUILO CHAMBER ANO AltRT
LIGHTS
ADO(Owm 1NTULOCX
AOOEO lltE-COATER
REMOVABlE IUN.O PLATFORM SENSOR
AOOIO ,OWO[R 0lUVIRY
HUMIDrtYCONTROl
HEAT PLATE CONTROL
HUMIOJTY S(NSOR
VALvt CONTROL N2 AND All
VIPATION ACTUATOllt FOIi RE -COA.TER
VUIRATION SfNSOR
AOOION2,1Jt
ADO[D DOOR SUtSOR
SCANNER STATUS
LASER STATUS
ADOEO POWDER TANK LEVEL Ethernet
P'fROMOERS OfF/ON .USS 100 KHZ (2)
CAMERAS OFF/ON AXIS •PO CARD RS 232/ 485
(CAM[lllA• UNKOR IOGIGE vtStON)
WATER LEAK SEN~

Figure 8. Control hardware architecture for SLM machine. Features for an advanced machine are
listed on the lower left-hand side. Without these features, the figure illustrates a basic SLM machine. I/O
refers to Digital/Analog Input/output.

22
 A schematic of the control system along with the hardware interface of actuators and
sensors is shown in Figure 8. At the heart of this hardware schematic is a controller, which can be
a PC, an IPC (Industrial PC) or a programmable logic controller (PLC). The various hardware
elements (actuators and sensors) of SLM machine are controlled via their respective interfaces.
This approach lends itself to an open and adaptable hardware architecture by allowing for
interfacing sensors and actuators from any OEM, if they support the provided interface. It should
be noted, that most of the components feature an Ethernet or USB-based interface, which can be
easily expanded using the features of the controller.

Open Machine Control Software

As a demonstration of the efficacy of the new protocols developed in the first program, the
research team developed a tool to interpret the SCAN file and stream commands over a
communication interface to run actual SLM machines per the protocol. The controller program,
called Machine Controller, is based on the LabVIEW™ software programming environment from
National Instruments. LabVIEW™ is a ubiquitous programming environment that greatly
enhances interoperability, which will be key to enabling software exchange between researchers
using differing physical hardware in the future. Members of the Lawrence Livermore National Lab
team developed the Machine Controller and ran it on an SLM machined purchased from
Aconity3D. The extended team demonstrated the use of this code on several commercial and
proprietary machines during program. In addition, the Machine Controller can run a virtual
machine, which is useful for visualizing scan paths, predicting build time, etc. Some specific
features of the controller:
• It reads SCAN protocol files and generates commands specific to hardware
• It uses a plugin architecture supporting multiple commercial machines
• It has built-in design flexibility to allow future plugins for new machines
• It has start, pause and abort controls
• It displays the progress of the build, any errors, and hardware specific information that
indicates the real-time quality of the build
• It runs unattended for periods exceeding
• It generates a log file
• It has controls for manual changes of machine parameters for test and setup
• It displays real-time process information received from the hardware
• It allows on-the-fly changes of parameters during the build process
The program can be launched from within LabVIEW™ as a “Virtual Interface,” or a precompiled
version can be run directly from Windows. Data from the input files flows to the machine through
various interfaces, as illustrated in Figure 9. The modules that transfer data between the interfaces
can be upgraded with new modules that support new machine actions, for instance diagnostics,
without disturbing old modules. An example LabVIEW™ user interface is shown in Figure 10. A
single human interface was developed with relatively easy transfer to other machines. Machine-

23
specific code for a new machine component, or an entirely new machine, can be easily
implemented by a user.

B Machine Machine_Run

Environment Environ_Run DLL interface to scanner

lnt32_t _cdacl lnitMirrors(Clustar •eadSize,double
Layer layer_Run * Ma1<Power, double *MaxSpeed,char Path ln[I);

G
lnt32_t _cdecl LoadPath(float XX(), float VY[), float
Power[], float Velocity(), int32_t Ion);
Scanner Scanner_Run lnt32_t _cded Mark(vold);
int32_t _ cdecl Markin1Status(int16_t ' Markin1);
lnt32_t _cdecl SetPower(doub1e Power);
Scanner Scanner_Run int32_t _ cdecl SetVelocity(dcuble Velocity);

I ------ I
New file formats Scanner Scanner_Run
I I
I Future I can be New machines can be
I Format I supported by Recoater Recoater_Run supported by writing a
I I
l - - - - -1 writing a driver DLL t hat conforms
converter which Layer Layer_Run to the i nterfaces of the
is dynamically ~-- ---,
I I
various interpreters. The
called based on New process
file extension '-----·
~-----, blocks can be
DLL can be written in any
I I languange, so a new
'-----• added by machine can be supported
writing an w ith no LabVIEW
interpreter programming
Process
Input File Converter VI Interpreter VI Shared library (DLL)
Blocks

Figure 9. Data Flow in the Machine Controller.

24
 •· •·
-·
--=·.1 -·
=·.•1
. .- •:

Figure 10. Example LabVIEW™ Virtual Interface. This custom control was developed at LLNL for
microsecond pulse laser machining incorporating real-time sensing, FPGA control, and galvanometer-
scanned laser beam motion.

Controller Demonstration
A research SLM system developed at Rensselaer is an exemplary platform for
demonstrating the versatility of the control architecture. The system is a combination of repurposed
and new components and is thought to be representative of what many institutions would be
capable of developing to advance research in additive manufacturing. Application of the open
controller architecture to this machine showcases its ability to integrate a variety of devices over
many interfaces. Implementation of these interfaces will facilitate the controller’s further
development and application to other machines, much as the open source community is expected
to do.
The major components of Rensselaer’s research SLM machine are shown in Figure 11 and
the assembled system is shown in Figure 12. The SLM process takes place inside a vacuum
chamber that was repurposed from another lab. This allows replacement of the atmosphere, as the
ambient air can be vacuumed out and back-filled with an inert gas, such as argon. The build plate
and powder supply are moved on piston-style axes, while the recoater blade is driven along linear
axes inside the chamber. A 400-watt fiber laser is used to provide energy to the process, which is
focused by an F-theta lens and steered onto the powder bed by means of a galvanometric scanner.
Sensors are used to monitor the machine and the process, which provide additional data to the
controller for use in real-time and in post-processing and analysis. Redundant safety features, such
as emergency stops and door interlocks, ensure the safety of the operator and bystanders during
operation

25
 The control cabinet has two main compartments (Figure 12). There are standard 19” rack
mounts, attachment rails, and a shelf for placing equipment. The top component contains a shelf
but can be modified to suit various needs. All the discrete power distribution and control
components can be mounted on DIN rails that are accessible from the rear of the cabinet. An
emergency stop button and machine status indicator lights are mounted on top of the cabinet. The
lower compartment can be locked to prevent unauthorized access to the laser and other sensitive
components.
A coin was chosen for the Rensselaer demonstration part. This coin has the GE Monogram
on one side, the Rensselaer seal on the other side, and the America Makes seal on the edge. The
scan file preparation (see Figure 13) followed the sequence described above for SCAN file
generation, and a resulting part, a coin, also shown in the figure.
Typical process trials are shown in Figure 14. “Bead-on-Plate” trials, where, the system is
exercised without metal powder are very common when debugging an SLM machine. The width
of the weld pool can be measured easily from the top down with a microscope. Often the sample
is cut, polished and placed under a microscope to measure the depth of the weld pool. Bead-on-
plate trials are critically important for setting the scanner and laser synchronization delay
parameters. Process parameter trials are extremely common. Here, samples are made at various
laser speeds, powers, hatch spacings, etc., to determine an optimal set for a given material. This
kind of parameter trial is exactly what the America Makes control system was designed to enable.

Power Meter Zoom Beam Expander Galvo Scan Head

Fused
Silica
Chamber
Window

~
_I
-;j!ll ;
Ju HI IJso·.::i.,
•
.. ··:o ~
--- -----(;)
..

Figure 11. Major components of the Rensselaer SLM system.

26
 Figure 12. Rensselaer's research SLM system. The mechanical components are housed in the
repurposed vacuum vessel to the left. The controller is housed in the cabinet to the right.

Layers 1-40
Supports {2mm)

Figure 13. Build File Preparation Sequence for Demonstration Coin. The final coin is shown on the
lower right.

27
 Figure 14. SLM Trials on the Rensselaer Machine. Bead on plate trials on the left, and process
parameter trials on the right.

Project #4: America Makes #3024

Acceleration of Large Area Additive Manufacturing (ALSAM)
Dec. 2018 – March 2022

The fourth program, ongoing, is a culmination of the prior three. During this program, a
commercially available Concept Laser M2 machine will be modified to allow operation using
either the OEM-supplied scan path generation and controls software or the America Makes open-
source controls system for multi-laser control. The resulting “M2-Open” system will be supplied
to the Air Force Research Lab, including the source code for all software, by the end of 2020. The
program overcomes research barriers by providing an open-architecture machine that runs open-
source software. Researchers will have access to all the parameters and scan path software
necessary to fully explore solution space. This capability will speed the maturity of additive
manufacturing processes and facilitate widespread adoption.
Control of the M2-Open is illustrated in Figure 15. When operated in “OEM-mode,” the
system will use commercially available scan path generation software, and the standard OEM-
supplied operating system known as CL WRX, with no changes. When operated in “open-mode,”
the system will require an input file conforming to the SCAN protocol, and the LabVIEW™
system controller will drive all functions of the machine including the laser and scanner. Penn
State is participating in the program by refining the spectroscopy-based sensors and analytics from
Program #2 for use on both lasers of the M2-Open.
A major contribution of this program will be to adapt and develop the SCAN protocol and
the open-controller to support multiple lasers. The updated SCAN protocol is designed to enable

28
customizable scan path generation routines, whether based on the open-source software described
above or 3rd party software. In addition, the synchronized multi-laser control will allow for the
development and evaluation of scan strategies optimized for various objectives, such as maximum
throughput or tailored thermal profiles. Several options for multi-laser build which will be added
enable either faster build rates [7] or more even control of heating along the build direction [8].
Designed experiments to address overhangs and surface roughness [9] [10] [11] will be enabled.
In addition, segment-by-segment control of power, speed and focus will be enabled to allow
delicate or frequently trouble-prone build areas to be fine-tuned [12] [5].
In addition to multi-laser control, the revised SCAN protocol will incorporate new features
such as build “styles", which describe entire parameter sets covering power, speed, focus and such.
This will enable the SCAN files to be greatly condensed, while still permitting optional segment-
by-segment control of build parameters. The LAYER and SCAN file protocols will be modified
to enable these new features.
Most of the machine operations are handled through secondary controls. For example, the
scanners controller (SCANLAB RTC5 units) are control cards for the laser scanners. These cards
sit in the OEM-supplied industrial PC which serves as the supervisory controller and are software
addressable from LabVIEW™. Standard motion of actuators, valves, gauges, and other items is
handled by an OEM-supplied PLC which is software-addressable from the supervisory controller
via LabVIEW™. To allow extensibility of the M2-Open to future possible enhancements, as set
of digital and analog I/O cards, also software-addressable through LabVIEW™, will be supplied
as RSTi EP Slice modules. A key concern is operator safety. The original M2 machine uses a
separate PLC to monitor all safety-related aspects of the machine including laser operation, door
status, oxygen level, etc. to prevent inadvertent dangerous operation by the operator. This Safety
PLC will be left unmodified in the M2-Open. The LabVIEW™ controller can read from the Safety
PLC but cannot write to it. This will ensure that all factory-installed safety features remain
uncompromised when the system is operated in either mode.

GE Concept Lase
Operating Software
M2
(Choose 1)

UDP / Controller
Ethernet Protocol

MODBUS / ,..- OEM-

Ethernet
Supplied
RSTT EP Slice Safety PLC
/0 Modules

29
 Figure 15. Concept Laser M2-Open. Items circled in red indicate the modifications to the controls
system to implemented during the ALSAM project.

Conclusions
The research community, which includes academia, national laboratories, industrial
research laboratories, and independent researchers have long voiced frustration with “closed”
OEM manufacturing equipment that hampers their ability to perform independent research beyond
the design intent of the original equipment manufacturer. This is true in many research fields,
though the series of research programs discussed in this paper addresses concerns restricted to the
SLM field. The programs discussed here have generated a set of software and hardware that allows
complete access to SLM machines through an “open” approach. This open approach includes
source code file generation and machine control, extensible hardware for easy system
modification, and advanced sensors for monitoring the process with the goal of improving the
SLM process.

30
 Acknowledgements
The four research programs discussed in this paper were sponsored by the United States
Air Force through America Makes, with contributions from the General Electric Aviation business.
America Makes is managed and operated by the National Center for Defense Manufacturing and
Machining, NCDMM. The authors gratefully acknowledge the support of these organizations.
At Rensselaer, special thanks to Dr. Steven Rock and his students for their work on
implementing the control system on the RPI DMLM machine.
Members involved in the ARL Penn State Open Protocol program included Alexander J.
Dunbar, Edward W. Reutzel, and Richard Martukanitz from ARL Pen State; Jared J. Blecher and
Neal Orringer from 3D Systems; George Levesque and Suresh Sundarraj from Honeywell; and
Sung S. Park from Northrop Grumman Corporation.
At America Makes, special thanks to John Wilczynski, Scott Crynock, and Rick Fowler
for their ongoing enthusiasm and support.

References

[1] S. Roychowdhury, W. T. Carter, M. J. Mahony, M. R. Ticker and D. S. Toledano, "An Open

Source Build Software and Controller Platform for Powder Bed Fusion Additive
Manufacturing," in Solid Freeform Fabrication Symposium, Austin TX, 2019.

[2] D. Lewis, A. Maniatty, J. Samuel, C. Carothers, S. Rock, S. Peters, J. Dolan, A. Chowdhury,

M. Rand, V. Titze, P. Miller and A. Waltzer, "Progress Towards a Validated Thermal
Material Simulation of the Additive Manufacturing Process in Inconel-718," in New York
Manufacturing Conference, Troy, NY, 2019.

[3] A. Ranellucci, "Slic3r.org," 2019. [Online]. Available: https://slic3r.org/.

[4] A. J. Dunbar, A. R. Nassar, E. W. Reutzel and J. .. Blecher, "A Real-time Communication

Architecture for Metal Powder Bed Fusion Additive Manufacturing," in Proceedings of the
27th Annual International Solid Freeform Fabrication Symposium, Austin, TX, 2016.

[5] A. J. Dunbar and A. R. Nassar, "Assessment of optical emission analysis for in-process
monitoring of powder bed fusion additive manufacturing," Virtual and Physical Prototyping,
vol. 13, no. 1, pp. 14-19, 2018.

[6] S. J. Rock, "Toward Accessible Metal Additive Manufacturing with Open Architecture
Controls," in ASM Spring Symposium, Niskayuna NY, 2017.

31
[7] M. Masoomi, S. M. Thompson and N. Shamsaei, "Quality part production via multi-laser
additive manufacturing," Manufacturing Letters, vol. 13, pp. 15-20, 2017.

[8] T. Heeling and K. Wegener, "Computational investigation of synchronized multibeam

strategies for the selective laser melting process," in 9th International Conference on
Photonic Technologies - LANE 2016, Furth, Germany, 2016.

[9] H. Chen, D. Gu, J. Xiong, Xia and Mujian, "Improving additive manufacturing processability
of hard-to-process overhanging structure by selective laser melting," Journal of Materials
Processing Technology, vol. 250, pp. 99-108, 2017.

[10] L. E. Criales, Y. M. Arisoy, B. Lane, S. Moylan, A. Donmez and T. Ozel, "Laser powder
bed fusion of nickel alloy 625: experimental investigations of effects of process parameters
on melt pool size and shape with spatter analysis," International Journal of Machine Tools
and Manufacture, vol. 121, pp. 22-36, 2017.

[11] L. Ma, J. T. Fong, B. Lane, S. P. Moylan, J. J. Filliben, N. A. Heckert and L. E. Levine,

"Using DOE in Finite Element Modeling to Identify Critical Variables in Laser Powder Bed
Fusion," in 2015 Annual International Solid Freeform Fabrication Symposium. An Additive
Manufacturing Conference, Austin, TX, 2016.

[12] J. C. Fox, S. P. Moyhlan and B. M. Lane, "Effect of Process Parameters on the Surface
Roughness of Overhanging Structures in Laser Powder Bed Fusion Additive
Manufacturing," in 3rd CIRP Conference on Surface Integrity (CIRP CRI), Charlotte, North
Carolina, 2016.

32