2. COMPUTER AIDED MANUFACTURING (CAM) IS OLD
▪ It’s been used for at least 40 years.
More depending on your definition
▪ Enhanced precision
▪ Reduced material consumption, better
material tracking
▪ Faster design-to-production cycle
▪ Automated customization
▪ Extremely repeatable
3. ADDITIVE MANUFACTURING IS JUST CAM
▪ Enhanced precision
▪ Reduced material consumption,
better material tracking
▪ Faster design-to-production
cycle
▪ Automated customization
▪ Extremely repeatable
5. WHY DO WE CARE?
▪ NASA (SpaceX, MoonEx, Virgin, etc) want 3D
printed rocket parts
▪ Boeing wants 3D printed jet parts
▪ Rolls-Royce wants 3D printed engine parts
▪ 70% reduction in part count
▪ 80% reduction in lead time
▪ 90% reduction in vendor lock in
▪ 100% reduction in being lame
6. WHY ISN’T THIS SOLVED? WE’RE IN SILICON VALLEY!
▪ Analysis without destruction is
hard (NDE)
▪ Analysis of highly detailed
structures is harder
▪ Analysis while it’s being
constructed is harderer
...but everyone wants it
7. HOW DO PEOPLE SOLVE THIS PROBLEM NOW?
▪ Know your model
▪ Know what your parameters are
▪ Know what your slicer should
generate
▪ Know what your printing should be
doing
▪ ...and watch it print.
8. ▪ Most prints cross multiple
communications barriers
▪ These barriers make tool
switching easier
▪ They make pipeline
knowledge harder
▪ QA and Repeatability rely on
full process pipeline
information
LET’S AUTOMATE THIS (1)
9. LET’S AUTOMATE THIS (2)
▪ Consolidate process
information
▪ Consistent API
▪ Historical information
▪ Multi-directional
communication
10. LET’S AUTOMATE THIS (3)
▪ Your eyes are
‘Monitor’
▪ Your brain is
‘Evaluation’
▪ Your experience is
input
12. ▪ Capture slicing engine output and create expected images
▪ Inject gcode to reposition head at opportune moments
▪ Process images and analyze for failures
HOW?
16. HUMAN
CONFIRMATION
▪ We’re at the early
stages
▪ False positives are
expensive
▪ The user knows best
▪ Bootstrapping the robot
revolution
17. THE REAL DEAL
▪ Authentise Monitor 2.0
▪ Bootstrap from FDM to binder jet to metal
▪ Observe more with more sensors
▪ Learn more with more data
▪ Operate on machine learning
19. ADDITIONAL SENSORS
▪ Good QA relies on triangulation of data sources
▪ Repeatability relies on constraining variables
▪ Sensors:
▪ Temperature
▪ Vibration
▪ Sound
▪ Non-visible spectrum
▪ Ultrasonics
▪ Holography/Shearography/Thermography?
20. MACHINE LEARNING (1)
▪ Machine learning now buys us feature (failure) recognition
▪ Machine learning in the future:
▪ Model analysis (80% of prints like that need thicker walls)
▪ Slicer parameter analysis (This printer/model will go better with
different infill)
21. MACHINE LEARNING (2)
▪ Printer parameter analysis (Your shell speed should be reduced)
▪ Ongoing print analysis (Looks like gears are slipping)
▪ Required printer maintenance (Your thermistor is going to fail soon)
▪ Printer selection (You’ve got a Fortus, use that for this)
22. OPERATE ON EXPERIENCE
▪ As we gain confidence
in the system we can
make it work for us
▪ Pausing/aborting prints
▪ Adjusting printer
parameters in situ
▪ Re-laying powder beds
▪ Clearing supply feeds
▪ Modulating power/binder
23. CLOSING THE LOOP
▪ AM’s greatest strength
cannot be forfeit to serve QA
needs
▪ What we need is an adaptive
process to qualify our
adaptive process
▪ Understanding our output
comes through
understanding our input
24. ADDITIVE MANUFACTURING IS CAM
▪ Enhanced precision
▪ Reduced material consumption, better
material tracking
▪ Faster design-to-production cycle
▪ Automated customization (of QA)
▪ Extremely repeatable