OpenLMD (http://openlmd.github.io/) is a set of software components provided to demonstrate last advances on laser processing control systems. Built on ROS (Robot Operating System), the modular approach of OpenLMD pursues a direct deployment of new algorithms beyond the state-of-the-art in real facilities, fixing common interoperability and standardization issues. Moreover, it takes advantage from open source and most advanced robotics, vision, and machine learning research.
1. Open Laser Metal Deposition
(OpenLMD)
Jorge Rodríguez-Araújo
AIMEN Technology Center, Porriño, Spain
Porriño, 7-12-2016
2. openlmd.github.io | jorge.rodriguez@aimen.es 2
Index
1. Motivation and Innovative Character
2. OpenLMD, modular architecture
3. ROS-based LMD cell integration
4. 3D geometrical monitoring
5. Multimodal monitoring
6. Off-line robot path planning
7. Real-time power control
8. Adaptive LMD path planning
9. Big data registration and analysis
10.Conclusions and future work
Index
4. openlmd.github.io | jorge.rodriguez@aimen.es 4
Motivation and Innovative Character
Promising additive manufacturing technique.
Parts are built up layer by layer directly from a 3D CAD model.
For repair and direct fabrication of pieces.
Near-net-shape (close to the final shape).
Manufacturing of large metallic parts.
The material is directly deposited on the previous surface.
Complex setup and adjustment of parameters.
Thermal heating accumulation produces geometrical distortions.
Distortions rise in poor dimensional accuracy and defects.
Traditional off-line process (with constant parameters)
becomes unsuccessful.
Laser Metal Deposition (LMD)
LMD Issues
5. openlmd.github.io | jorge.rodriguez@aimen.es 5
Motivation and Innovative Character
There are a lot of industrial robotized laser cells.
Empower robotized laser cells for effective AM.
Retrofit current industrial facilities.
Apply state of the art robotic
software solutions.
Motivation
Robotized
Cladding Cell
Motion
Controller
Main Controller
Off-line
Path Planning
6-Axis
Robot
Laser
Powder
Feeder
Power
Controller
Flow
Controller
Innovation
6. openlmd.github.io | jorge.rodriguez@aimen.es 6
Multiprocessing architecture based on message publishing
Multi-node and multi-machine
Synchronized multimodal data acquisition
Multiple high speed image sensors
Process equipment (e.g. robot, laser) and data
RT-control, cloud storage, data analysis, and visualization
Common timestamp
High bandwidth multimodal data storage and analysis
Thinked for deep learning algorithms
Open Laser Metal Deposition
ROS-based architecture
8. openlmd.github.io | jorge.rodriguez@aimen.es 8
OpenLMD, modular architecure
CONCEPT
Open-source solution for on-line multimodal monitoring and control of LMD
Modular set of software components. Built on ROS (Robot Operating System)
Interoperability and standardization
Robotics, machine vision, embedded control, machine learning
ROS-based LMD cell integration
Integration of ABB LMD robotized cell based on ROS.
3D Geometrical Monitoring
3D on-line monitoring (point cloud) for LMD robotized cells.
Multimodal monitoring
Image-based multimodal monitoring for Laser applications.
Off-line Robot Path Planning
Off-line path planning for robotized LMD automation.
Real-Time Process Control
Image-based asynchronous RT close-loop control for LMD systems.
Adaptive LMD Process Planning
Adaptive path planning for an automatic repair of large and complex metal parts.
Big Data Registration and analysis
Big data registration for LMD adaptive parametric control (for Cloud Computing and
Deep Learning)
9. openlmd.github.io | jorge.rodriguez@aimen.es 9
ROS-based modular laser cell integration
The PC integrates the interface and modules to command the robot (ROS)
The robot controls all the cell elements
ROS components:
Robot state publisher
Robot command server
ROS-based LMD cell integration
Powder
Feeder
Fiber Laser
6-Axis Industrial Robot
PC
Controller
Cladding
Head
ROS-Driver
(ABB Rapid)
Geometrical
Cell Description
(URDF)
STATE
PUBLISHER
Laser Source
(slave)
Powder Feeder
(slave)
COMMAND
SERVER
Power
Speed
Powder flow
Motion path
States
Commands
ROBOT
Process
parameters
10. openlmd.github.io | jorge.rodriguez@aimen.es 10
On-line 3D scanning
Real –time point cloud registration
Actual metric measurements (mm)
Direct acquisition in robot coordinates
3D geometrical monitoring
Industrial Robotic Laser Cell
ROBOT
ROS-DRIVER
CAMERA
IDS-DRIVER
State Publisher Peak Finder
Robot Pose
Tool-Camera
Laser
TriangulationCalibration
3D ProfileCamera Pose
3D Point Cloud
Working Cell Coordinate
Point Cloud Reconstruction
11. openlmd.github.io | jorge.rodriguez@aimen.es 11
Multimodal monitoring
Multimodal Cladding Head
3D System
Tachyon
MWIR
NIR
Multimodal monitoring approach
Coaxial SWIR/MWIR images (thermal monitoring): NIT microcore (1000fps) [1-3um]
Coaxial NIR images (surface monitoring): CMOS camera (100fps) [830-880nm]
Off-axis 3D system: on-line 3D point cloud scanning (50fps)
MWIR+NIR
Multispectral
Imaging
LMD Cell Virtualization
12. openlmd.github.io | jorge.rodriguez@aimen.es 12
Off-line robot path planning
Robot Routine
(ABB rapid)
Intuitive robot programming
Automatic generation of robot trajectories
User friendly interface with high level of abstraction
CAD-based off-line programming (no robot programming skills needed)
3D Part Visualization
13. openlmd.github.io | jorge.rodriguez@aimen.es 13
Melt pool
geometry
PI controller
(Kp, Ki)
CLADDING
process
width
powereSP
PV
MicroCore
Close-Loop Control
Real-Time power control
Without control With control
On-line asynchronous laser power control
Closed-loop control
Embedded vision
RT Control Interface
14. openlmd.github.io | jorge.rodriguez@aimen.es 14
Adapts the path to the real geometry
Automatic repair of large parts
On-line geometrical monitoring
Adaptive path planning
Geometrical control
Adaptive LMD path planning
Robotized Laser Cell
Track MeasurementPath Planning
3D Model
Layer Planning Layer Measurement
STL Generation
On-line
3D Filtering
Initialization
(setup)
Scan layer
Depth map
Target
Depth map
Disparity
Data
Layer path
planning
Layer Path Planning
(geometrical control)
Laser Cell
supervisor
Robotized
Cell
0 Finished
Repair Job
Adaptive path planning
15. openlmd.github.io | jorge.rodriguez@aimen.es 15
Cyber Physical registration
High throughput (28MB/s) (NIR + MWIR + 3D point cloud+ Robot)
Deep learning capabilities
Data management and analysis (DataFrames)
Feature
extraction
Integration of
temporal data
Iterative
adjustment
Quality
diagnosis and
reconfiguration
Process
parameters
LMD
Big data registration and analysis
60 GB/h
Cyber Physical Sytem
17. openlmd.github.io | jorge.rodriguez@aimen.es 17
Data acquisition
Spatial reference sistem and temporaly synchronized
Annalysis
Deep learning: features extration
Actuation
Real-time laser power control
Adaptive path planning
Embedded vision and control systems
Robot
Pose
Process speed
3D geometry
Point cloud (<0.5mm)
Multimodal: SWIR/MWIR-NIR
2D melt pool geometry
Thermal distribution
and texture
Reconfigurable
Conclusions
Conclusions
Current lines
Modular and
reconfigurable
Interoperability
Large parts
Low-cost
solution
Scalability
18. AIMEN – Central y Laboratorios
c/ Relva 27 A
36410 – O PORRIÑO (Pontevedra)
Telf.+34 986 344 000 – Fax. +34 986 337 302
Thank you for your
attention
Jorge Rodríguez-Araújo | Research Engineer
Ph +34 986 344 000 | jorge.rodriguez@aimen.es
www.aimen.es | aimen@aimen.es